Simple Moving Average Beispiel Prognose

Businessballs hat zusammen mit Accipio mdash ein Institut für Leadership and Management (ILM) und das CMI-Zentrum (m / w) eingerichtet, um kostenlose audiovisuelle interaktive eLearning-Module anzubieten, die mit international anerkannten Qualifikationen (ILM oder CMI) abgestimmt sind. Erlernen von Lernpunkten für jede Führungs-und Management-eModule, und gewinnen Sie ein Level 3 Award, Zertifikat oder Diplom, sobald Sie mit der Vergabeorganisation (über Accipio registriert haben), gesichert genug Lernpunkten und verabschiedete die Aufgaben. Akkreditierungsgebühren gelten. Klicken Sie hier, um auf die eLeadership Academy zuzugreifen. Business-Pläne und Marketing-Strategie Kostenlose Business-Planung und Marketing-Tipps, Beispiele, Beispiele und Tools - wie man einen Business-Plan zu schreiben, Techniken für das Schreiben einer Marketing-Strategie, strategische Business-Pläne und Verkaufspläne Hier sind Tipps, Beispiele, Techniken, Werkzeuge und einen Prozess Um Geschäftspläne zu schreiben, um effektive Ergebnisse zu erzielen. Dieser kostenlose Online-Leitfaden erklärt, wie man eine Marketing-oder Business-Strategie, einen grundlegenden Business-Plan und einen Verkaufsplan, mit kostenlosen Vorlagen, Tools und Beispiele, wie SWOT-Analyse zu schreiben. PEST Analyse. Die Ansoff Matrix und die Boston Matrix. Getrennt bietet der Marketing-Führer eine spezifischere Erklärung und Theorien und Werkzeuge für Marketing-Strategie und Marketing-Planung, einschließlich Techniken und Tipps für Werbung, Öffentlichkeitsarbeit (PR), Presse - und Medienwerbung, Verkaufsabfrage Lead-Generierung, Werbetexten, Internet und Website Marketing, etc. Die Sales-Schulung Führer bietet detaillierte Theorien und Methoden über Vertriebsplanung und Verkauf, die sich auf Kaltanrufe und Verhandlung Fähigkeiten und Techniken. Insbesondere im Zusammenhang mit dem Verkauf. Manchmal verwenden die Menschen den Begriff Businessplan, wenn sie sich auf ein Projekt beziehen. Es kann oder auch nicht angemessen sein, den Begriff Geschäftsplanung für ein Projekt zu verwenden. Einige Projekte sind sehr umfangreich und entsprechen einer autonomen (unabhängigen) Geschäftstätigkeit. In diesem Fall ist ein Businessplan völlig angemessen. Andere Projekte sind kleiner, vielleicht auf interne Veränderung oder Entwicklung beschränkt und sind weniger wahrscheinlich, einen konventionellen Business-Plan zu erfordern, und sind durchaus adäquat geplant und verwaltet über Projektmanagement-Methoden. Business-Terminologie kann verwirrend, weil viel von ihm verwendet wird sehr locker, und kann verschiedene Dinge bedeuten. Hier ist ein Weg, um es besser zu verstehen: Business-Planung Terminologie .. Terminologie in der Unternehmensplanung wird oft sehr lose verwendet. Wenn Menschen reden und schreiben über Business-Planung verschiedene Begriffe können die gleiche Sache, und ein einziger Begriff kann verschiedene Dinge bedeuten. Der Begriff Business-Planung umfasst alle Arten von verschiedenen Plänen innerhalb eines Unternehmens, oder möglicherweise innerhalb einer nicht-kommerziellen Organisation. Die Worte Strategie und strategische entstehen oft in das Thema der Business-Planung, obwohl es keinen wirklichen Unterschied zwischen einem Business-Plan und einem strategischen Business-Plan. Jeder Businessplan ist wohl strategisch. Jeder, der in die Planung involviert ist, nimmt wohl einen strategischen Ansatz an. Die meisten Geschäfte und Pläne werden hauptsächlich durch Marktbedürfnisse und - ziele bestimmt oder bestimmt. Dies gilt in zunehmendem Maße für viele nichtkommerzielle Aktivitäten (staatliche Dienstleistungen, Bildung, Gesundheitswesen, Wohltätigkeitsorganisationen usw.), deren Planungsprozesse auch als Unternehmensplanung bezeichnet werden können, auch wenn es sich dabei nicht um Unternehmen handelt, wie wir es normalerweise vorstellen. In solchen nicht kommerziellen Organisationen kann die Unternehmensplanung stattdessen Organisationsplanung oder operative Planung oder jährliche Planung oder einfach planen. Im Wesentlichen alle diese Begriffe bedeuten die gleiche, und zunehmend die Tendenz ist für die Unternehmensplanung zu einem generischen (allgemeinen) Begriff, um auf sie beziehen. Ich sollte klarstellen, dass die Finanzierung natürlich ein wichtiger und unvermeidbarer Aspekt der geschäftlichen und organisatorischen Aktivitäten ist, aber in Bezug auf die Planung, Finanzierung ist ein begrenzender oder förderlicher Faktor Finanzierung ist ein Mittel zum Zweck oder eine Beschränkung Finanzierung an sich ist keine Grundlage Für Wachstum oder Strategie. MärkteKunden, Produktentwicklung und Vertrieb, bieten die einzige wahre Grundlage für Unternehmen zu definieren Richtung, Entwicklung, Wachstum, etc. und damit Business-Strategie und Planung. Business-Planung beginnt immer mit dem grundlegenden Ziel oder die Notwendigkeit, Produkte oder Dienstleistungen an Kunden - auch als Markt oder Marktplatz. Konsequenterweise sehen die Business-Pläne zunächst nach außen, auf einem Markt, bevor sie nach innen schauen, auf Finanzen und Produktion usw. Das bedeutet, dass die meisten Business-Pläne von Marketing getrieben werden, da Marketing die Funktion ist, die Marktchancen und - bedürfnisse adressiert Zu erfüllen. Marketing in diesem Sinne wird auch als Marketing-Strategie oder breitere Geschäftsstrategie. In vielen einfachen, kleinen und alten traditionellen Unternehmen, Marketing ist oft gesehen, statt Verkauf oder Verkauf (in der Regel, weil in diesen Geschäften Verkauf ist die einzige Marketing-Aktivität), in diesem Fall ein Absatzplan kann der Hauptantrieb der Strategie und die sein Geschäftsplan. Viele Leute verwenden die Wörter Verkauf oder Verkauf und Marketing, um die gleiche Sache - im Grunde den Verkauf von Produkten oder Dienstleistungen an Kunden, im weitesten Sinne. In der Tat bezieht sich das Marketing auf viel breitere Fragen als Vertrieb und Verkauf. Marketing umfasst die strategische Planung eines Unternehmens (oder eines anderen organisatorischen Anbieters) bis hin zu allen Aspekten des Kundenengagements, einschließlich Marktforschung, Produktentwicklung, Branding, Werbung und Verkaufsförderung, Verkaufsmethoden, Kundenservice und Erweiterung auf die Akquisition oder Entwicklung von Neue Unternehmen. Verkauf oder Verkauf ist eine Tätigkeit im Marketing, die sich auf die Methoden und Prozesse der Kommunikation und Vereinbarung und Abschluss der Transaktion (Verkauf) mit dem Kunden. Angesichts all dessen ist es hoffentlich leichter verständlich, warum die Unternehmensplanung je nach Rolle oder Standpunkt oder der Abteilung, in der sie tätig sind, auf vielfältige Weise vermittelt werden kann, zum Beispiel als Absatzplanung, Marketingplanung und strategische Planung , Etc. und dass alle diese Begriffe etwas unterschiedliche Dinge bedeuten könnten, je nach der Situation. Wenn es eine technisch korrekte Definition der Unternehmensplanung gibt, dann können wir am besten sagen, dass die Unternehmensplanung sich auf den Plan der Gesamtorganisation oder auf eine Einheit oder Abteilung innerhalb einer Organisation bezieht, die für einen Handel oder Gewinn verantwortlich ist. Ein Businessplan enthält technisch die einzelnen Pläne für die verschiedenen Funktionen innerhalb des gesamten Vorhabens, die jeweils über eigene, detaillierte Businesspläne verfügen, die als Businesspläne bezeichnet werden können Als Marketingplan, Verkaufsplan, Produktionsplan, Finanzplan, etc. Zusätzliche Hilfe in Bezug auf Terminologie wird durch die unternehmerischen Planungsdefinitionen angeboten. Weitere Definitionen und Erläuterungen finden Sie im Business-Glossar. Und in den kürzeren Glossaren der Vertriebs - und Marketingabteilungen. Terminologie wird weiter erklärt werden, um die Bedeutung und Vermeidung von Verwirrung in diesem Artikel zu klären. Einführung Approach richtig, schriftlich Business-Pläne und Marketing-Strategie ist in der Regel einfacher als erste scheint. Business-Planung kann komplex und erschreckend erscheinen, aber meistens ist es gesunder Menschenverstand. Marketingstrategie, die oft die Ziele und Gestalt eines Businessplans vorantreibt, ist meist auch gesunder Menschenverstand. Business-Pläne und die Strategie, die sie antreibt, basieren auf Logik oder Ursache und Wirkung. Ich möchte ein bestimmtes Ergebnis erzielen - also was wird dazu führen, dass dies geschieht Selbst der größte Business-Plan ist effektiv auf eine Sammlung von vielen Ursachen und Auswirkungen gebaut. Ein schriftlicher Businessplan liefert die Erzählung (Erklärung) der Zahlen, die in einer Kalkulationstabelle enthalten sind. Wenn wir viele Zahlen in einer Computerkalkulation sehen, können wir dies vergessen, aber die Zahlen sind nur ein Spiegelbild von Skalierung und Detail, und von computergesteuerten Berechnungen und Modellierungen usw. In der Tat oft, wenn wir mit einer komplexen Planung Spreadsheet mit Tausenden konfrontiert sind Der Zahlen, was wir tatsächlich angeboten wird, ist ein fertiges Planungswerkzeug. In vielen Fällen, in denen die Unternehmensplanung eine Fortsetzung einer laufenden Situation ist, können die erschreckendsten Kalkulationstabellen eine sehr einfache Vorlage für zukünftige Pläne bereitstellen, besonders mit ein wenig Hilfe von einem Kollegen in der Acciiced-Abteilung, der versteht, wie alles funktioniert. Ironischerweise ist ein leeres Blatt Papier - also ein Neugründungsstart - meist ein viel schwierigerer Ausgangspunkt. Generell ist es schwieriger, einen Businessplan für ein Start-up-Geschäft (ein Neugeschäft) zu schreiben, als für ein bestehendes Geschäft. Dies liegt daran, ein bestehendes Geschäft hat in der Regel EDV-Aufzeichnungen über die Ergebnisse der vergangenen Aktivitäten und Handel (in der Regel Konten genannt). Kalkulationstabellen sind in der Regel mit Vorjahren Pläne und tatsächliche Ergebnisse, die als Vorlage verwendet werden können, auf denen neue Pläne können leicht überlagert werden. Das Schreiben eines neuen Geschäftsplans für die Fortsetzung oder Entwicklung einer solchen bestehenden Situation macht es offensichtlich möglich, dass ein Großteil der Planung auf bestehenden Zahlen, Kennzahlen, Statistiken usw. basiert. Neue Unternehmensgründungssituationen haben naturgemäß keine vorherigen Ergebnisse, So dass wir oft auf diese Art der Planung als beginnend mit einem leeren Blatt Papier beziehen. Neue Unternehmensgründungen - vor allem, wenn Sie Eigentümer oder Unternehmer sind - stellen in mancher Hinsicht größere Planungsherausforderungen dar, weil wir keine Vorreiter haben, um als Leitfaden zu fungieren, aber im Übrigen bieten sie wunderbare Möglichkeiten, echt innovative und spannende Gründungsprinzipien zu schaffen - Ihre eigene Firmenphilosophie, auf der Ihre Pläne aufgebaut und entwickelt werden können. Auf dieser Seite finden Sie konkrete Anleitungen für Existenzgründer. Siehe die einfachen Business-Start-up-Prinzipien. Abhängig von den Einschränkungen bei der Planung bestehender fortlaufender Geschäftsaktivitäten sind die Prinzipien für die Inbetriebnahme und die bestehende Unternehmensplanung sehr ähnlich. Sein im Wesentlichen Ursache-und Wirkung, und mit dem Computer, um die Zahlen zu berechnen. Eine etwas detailliertere Version befindet sich auf der schnellen Businessplanungsseite. Und beginnt mit Um persönliche Ausrichtung und Veränderung zu erforschen (z. B. für die frühzeitige Planung einer selbständigen Erwerbstätigkeit oder für Neugründungen), sehen Sie sich die Leidenschaft und die Vorlage auf der Seite zur Teambuilding-Übung an. Siehe auch die einfachen Hinweise zum Starten Ihres eigenen Unternehmens. Die in gewissem Maße auch gelten, wenn Sie eine neue Geschäftsinitiative oder Entwicklung in einer anderen Organisation als neuen Geschäftsentwicklungsmanager starten oder eine ähnliche Rolle. Heres eine kostenlose Gewinn-und Verlustrechnung Kalkulationstabelle Template-Tool (xls) für die Einbeziehung dieser Faktoren und Finanzen in eine formale phased Geschäft Handelsplan, die auch als Business-Prognose-und Reporting-Tool dient auch. Passen Sie es an Ihre Zwecke an. Dieses Planbeispiel ist auch als PDF erhältlich, siehe das Beispiel für die Erstellung eines Businessplans für das Unternehmen (Pampl). Die Zahlen könnten alles sein: zehnmal weniger, zehnmal mehr, hundertmal mehr - das Prinzip ist dasselbe. Zum Ende dieses Artikels gibt es auch ein einfaches Template Framework für eine Machbarkeitsstudie oder Rechtfertigung Bericht. Die erforderlich sind, um Finanzierung, Genehmigung oder Genehmigung für die Aufnahme eines Projekts oder die Fortsetzung eines Projekts oder einer Gruppe in einer kommerziellen oder freiwilligen Situation zu erhalten. Wenn Sie beginnen ein neues Unternehmen finden Sie auch die Tipps und Informationen über den Kauf eines Franchise-Geschäft hilfreich sein, da sie viele grundlegende Punkte über die Wahl der Geschäftstätigkeit und frühzeitige Planung umfassen. (Hinweis: Einige britische und US-englische Schreibweisen unterscheiden sich, z. B. Organisationsorganisation, colourcolor. Bei Verwendung dieser Materialien passen Sie die Rechtschreibung an Ihre Situation an.), Wie strategische Marketingpläne, Businesspläne und Verkaufspläne zu schreiben Die Menschen verwenden verschiedene Begriffe Die sich auf den Geschäftsplanungsprozess - Geschäftspläne, Geschäftsstrategie, Marketingstrategie, strategische Unternehmensplanung und Absatzplanung - beziehen, decken alle die gleichen Grundprinzipien ab. Im Hinblick auf die Unternehmensplanung oder Strategieentwicklung Aufgabe ist es wichtig, genau zu klären, was erforderlich ist: klären, was getan werden muss, anstatt das Ziel von der Beschreibung an sie - die Begriffe sind verwirrt und bedeuten verschiedene Dinge für verschiedene Menschen. Youll sehen Sie von den Definitionen unten, wie flexibel diese Geschäftsplanungausdrücke sind. Begriffsdefinitionen eines Planes - eine Absichtserklärung - eine berechnete Absicht, Anstrengung und Ressource zu organisieren, um ein Ergebnis zu erzielen - in diesem Zusammenhang ein Plan in schriftlicher Form, bestehend aus Erläuterung, Begründung und relevanten numerischen und finanzstatistischen Daten. In einem geschäftlichen Kontext sind Pläne numerischer Daten - Kosten und Erlöse - in der Regel über mindestens ein Wirtschaftsjahr geplant, aufgeschlüsselt wöchentlich, monatlich vierteljährlich und kumulativ. Ein Unternehmen - eine Tätigkeit oder ein Unternehmen, unabhängig von Größe und Autonomie, die in einer Tätigkeit tätig ist, in der Regel die Bereitstellung von Produkten und Dienstleistungen, um kommerzielle Gewinne zu produzieren, die sich auf nicht kommerzielle Organisationen, deren Ziel kann oder auch nicht profitieren (daher Warum Schulen des öffentlichen Dienstes und Krankenhäuser in diesem Zusammenhang als Unternehmen bezeichnet werden). Business-Plan - das ist nun zu Recht ein sehr allgemeiner und flexibler Begriff, der für die geplanten Aktivitäten und Ziele eines Unternehmens, einer einzelnen Gruppe oder Organisation gilt, bei dem die Ergebnisse in Ergebnisse umgesetzt werden. Zum Beispiel: ein kleines Unternehmen eine große Firma eine Ecke Geschäft ein lokales Fenster-Reinigung Geschäft ein regionales Geschäft ein Multi-Millionen-Pfund multinationalen Unternehmen eine Wohltätigkeitsorganisation eine Schule ein Krankenhaus ein Gemeinderat eine Behörde oder Abteilung ein Joint-Venture ein Projekt Innerhalb eines Unternehmens oder einer Abteilung ein Geschäftsbereich, eine Abteilung oder eine Abteilung innerhalb einer anderen Organisation oder eines Unternehmens, ein Profit-Center oder eine Kostenstelle innerhalb einer Organisation oder eines Unternehmens die Verantwortung eines Teams oder einer Gruppe oder einer Person. Die Unternehmenseinheit könnte auch eine Neugründung, eine neue Geschäftsentwicklung innerhalb einer bestehenden Organisation, ein neues Joint-Venture oder jedes neue Organisations - oder Geschäftsprojekt sein, das darauf abzielt, Maßnahmen in Ergebnisse umzusetzen. Das Ausmaß, in dem ein Unternehmensplan Kosten und Gemeinkosten beinhaltet (z. B. Produktion, Forschung und Entwicklung, Lager, Lagerung, Transport, Verteilung, Verschwendung, Schrumpfung, Zentrale, Schulung, Das Geschäft und den Zweck des Plans. Große Executive-Level-Business-Pläne schauen daher eher wie eine prädiktive Gewinn-und Verlustrechnung, vollständig bis unten aufgeführt. Geschäftspläne, die auf Geschäfts - oder Abteilsebene geschrieben werden, enthalten in der Regel keine Finanzdaten außerhalb des betreffenden Bereichs. Die meisten Business-Pläne sind in Wirklichkeit Vertriebs-Pläne oder Marketing-Pläne oder Abteilungspläne, die die Haupt-Bias dieser Anleitung bilden. Strategie - ursprünglich ein militärischer Begriff, in einem unternehmensplanerischen Kontext, der strategisch-strategisch ist, bedeutet, warum und wie der Plan funktionieren wird. In Bezug auf alle Einflussfaktoren auf das Unternehmen und die Tätigkeit, insbesondere die Konkurrenten (also die Verwendung eines militärischen Kampfes), Kunden und Demographie, Technologie und Kommunikation. Marketing - von vielen geglaubt, dass die gleiche wie Werbung oder Verkaufsförderung bedeuten, bedeutet Marketing tatsächlich und deckt alles von der Unternehmenskultur und Positionierung, durch Marktforschung, neue Businessproduct Entwicklung, Werbung und Promotion, PR (Public Press Relations), und wohl alle der Vertriebsfunktionen. Marketing ist der Prozess, durch den ein Unternehmen entscheidet, was es verkaufen wird, an wen, wann und wie und dann tut es. Marketing-Plan - logisch ein Plan, die Details, was ein Unternehmen verkaufen wird, an wen, wann und wie implizit einschließlich der Businessmarketing-Strategie. Das Ausmaß, in dem finanzielle und kommerzielle numerische Daten enthalten sind, hängt von den Bedürfnissen des Unternehmens ab. Das Ausmaß, in dem diese Details der Umsatzplan auch abhängig von den Bedürfnissen des Unternehmens. Vertrieb - die Transaktionen zwischen dem Unternehmen und seinen Kunden, wobei Dienstleistungen und Produkte gegen Entgelt zur Verfügung gestellt werden. Sales (Sales Departmentsales Team) beschreibt auch die Aktivitäten und Ressourcen, die diesen Prozess ermöglichen, und Umsatz beschreibt auch die Umsatzerlöse, die das Geschäft aus der Vertriebsaktivitäten stammt. Verkaufsplan - ein Plan, der beschreibt, quantifiziert und schrittweise im Laufe der Zeit, wie die Verkäufe gemacht werden und an wen. Einige Organisationen interpretieren dies als ein Business-Plan oder ein Marketing-Plan. Business-Strategie - siehe Strategie - es ist das gleiche. Marketing-Strategie - siehe Strategie - es ist das gleiche. Dienstleistungsvertrag - ein formelles Dokument, das in der Regel vom Lieferanten erstellt wird, mit dem das Handelsabkommen mit dem Kunden vereinbart wird. Siehe Abschnitt Serviceverträge und Handelsverträge. Strategische Business-Plan - siehe Strategie und Business-Plan - es ist ein Business-Plan mit strategischen Treibern (die eigentlich alle Business-Pläne sollte). Strategische Geschäftsplanung - Entwicklung und Erstellung eines strategischen Businessplans. Philosophie, Werte, Ethik, Vision - dies sind die Grundlagen der Unternehmensplanung und bestimmen den Geist und die Integrität des Unternehmens oder der Organisation - finden Sie in der Anleitung, wie philosophische und ethische Faktoren in den Planungsprozess passen. Und auch die Grundsätze und Materialien in Bezug auf Corporate Responsibility und ethische Führung. Sie können sehen, dass viele dieser Ausdrücke austauschbar sind, also ist es wichtig, zu klären, was geplant werden muss, anstatt eine Bedeutung von dem Namen anzunehmen, der der Aufgabe gegeben wird. Allerdings können die hier erläuterten Prinzipien auf Geschäftspläne aller Art angewendet werden. Business-Pläne werden oft als unterschiedliche Namen - vor allem von Führungskräften und Direktoren delegieren eine Planung, die sie nicht verstehen, gut genug, um zu erklären. Zum Beispiel: Verkaufspläne, Betriebspläne, organisatorische Organisationspläne, Marketingpläne, Marketingstrategiepläne, strategische Geschäftspläne, Abteilungsplanpläne usw. Typischerweise spiegeln diese Namen die Abteilung wider, die die Planung durchführt Dokument ist weitgehend ähnlich. Weitere nützliche und relevante Business-Planung Definitionen sind in der Business-Wörterbuch das Verkaufs - und Verkauf-Glossar einige sind auch in den finanziellen Begriffen Glossar. Und mehr - vor allem für die Ausbildung - sind in der Business-und Training Akronyme Auflistung. Die auch amüsante Lichtrelief bietet, wenn diese Unternehmensplanung ein wenig trocken wird (seien Sie gewarnt, die Akronyme-Listen enthalten einige Erwachsene Inhalt). Wenn Sie einen Geschäfts - oder Betriebsplan schreiben, denken Sie daran. Eine nützliche erste Regel der Unternehmensplanung ist zu entscheiden, was Sie tatsächlich versuchen, zu erreichen und immer daran zu erinnern. Schreiben Sie Ihr Ziel groß als eine ständige Erinnerung an sich selbst, und an alle anderen beteiligt. Ihr zentrales Ziel sichtbar zu machen hilft Ihnen, die Ablenkungen und Verzerrungen zu minimieren, die häufig während des Planungsprozesses entstehen. Eine zunehmend wichtige und vielleicht zweite Regel der Unternehmensplanung ist es, eine starke ethische Philosophie zu Beginn Ihrer Planung zu etablieren. Von Anfang an ist dies eine wichtige Referenz für Entscheidungsfindung und Strategie. Ein starker klarer ethischer Code kommuniziert Ihre Werte mit Mitarbeitern, Kunden, Lieferanten und schafft eine einfache, konsistente Basis für Operationen, die herkömmliche Finanzsysteme, Prozesse, Systeme und sogar Menschen nicht adressieren. Es ist sehr schwierig, ethische Grundsätze später in ein Unternehmen einzuführen, vor allem bei der Umsetzung von Umsetzungsschwierigkeiten und mehr, wenn Probleme in Bezug auf Integrität, Ehrlichkeit, Corporate Responsibility, Vertrauen, Governance usw. auftreten. Jede davon kann massive Auswirkungen auf Beziehungen haben Und Reputation. Siehe soziale Verantwortung der Unternehmen und Ethik und die psychologischen Vertrag. Es ist leicht, Fragen der Ethik und Corporate Responsibility anzugehen, wenn Sie der Besitzer eines neuen Unternehmens sind. Es ist schwieriger, wenn Sie ein Manager in jemand elses Firma oder ein großes Unternehmen sind. Dennoch sind Ethik und Unternehmensverantwortung bei der Planung von großer Bedeutung, und es kann eine starke Rechtfertigung für ihre richtige Berücksichtigung geleistet werden. Es gibt jetzt viele neue Beispiele von Unternehmen - in der Tat ganze Volkswirtschaften und Regierungen -, die aufgrund der schlechten Rücksicht auf ethische Erwägungen gescheitert sind. Die Welt verändert sich und das Lernen, langsam, aber es ist, und jeder ignorieren Ethik in der Planung heute tut dies auf eigene Gefahr. Eine dritte wesentliche Voraussetzung für Business-Pläne ist der Return on Investment. Oder für öffentliche Dienstleistungen und gemeinnützige Organisationen: effektive Nutzung von Investitionen und Ressourcen, die über eine einfache Kostenkontrolle hinausgehen. Für die überwiegende Mehrheit der Organisationen, ob Unternehmen, öffentliche Dienstleistungen, gemeinnützige Stiftungen und Wohltätigkeitsorganisationen, müssen alle Organisationen finanziell wirksam sein, was sie tun, sonst werden sie nicht mehr funktionieren. Letztendlich - unabhängig von der Organisation und den Zielen - ist die finanzielle Lebensfähigkeit notwendig, um jede organisierte Aktivität zu unterstützen. Während seine wesentliche ethische und sozial verantwortliche Aspekte der organisatorischen Ziele zu verwalten. Diese müssen eine angemessene Rentabilität (oder in weniger traditionellen und gemeinnützigen Unternehmen, müssen die effektive Nutzung von Investitionen und Ressourcen entsprechend den finanziellen Anforderungen der jeweiligen Organisation zu ermöglichen). Denken Sie an die Notwendigkeit der finanziellen Rentabilität ist von entscheidender Bedeutung auch, weil Unternehmensplanung ist oft getan - zu Recht - um etwas Neues und Besonderes zu erreichen. Dies neigt dazu, das Denken auf Kreativität, Innovation, Ehrgeiz, Qualität, Exzellenz, vielleicht sogar soziale Güter, etc., die leicht ablenken kann die Planung weg von der grundlegenden Notwendigkeit, finanziell tragfähig - und entscheidend nicht zu einem Verlust zu konzentrieren. Indem wir die Investitionsrentabilität als lebenswichtiges Planungskonzept behandeln, erhöhen wir die Wahrscheinlichkeit, dass die Pläne lebensfähig und daher nachhaltig sind. Return on Investment ist jedoch ein variables Merkmal der Unternehmensplanung. Es ist flexibel nach der Art des Unternehmens, seinen Hauptzweck und Philosophie. In einem herkömmlichen, gewinnorientierten Unternehmen ist der Return on Investment (mit einer optimalen Rate) typischerweise ein starker strategischer Treiber für lokale Planung und Entscheidungen und implizit auch eine grundlegende Anforderung des Unternehmens als Ganzes. Auf der anderen Seite, in einem Unternehmen oder einer Organisation weniger konzentriert auf Shareholder Belohnung, wie eine öffentliche Dienste Vertrauen oder Wohltätigkeitsorganisation oder ein soziales Unternehmen oder Genossenschaft, Return on Investment (mit einem relativ niedrigeren Satz), kann eine Anforderung einfach sein, zu erhalten Nach den Zielen des Unternehmens. Im ersten Beispiel ist der Return on Investment das Ziel im zweiten Beispiel, die Rendite auf die Investition ermöglicht es, ein anderes höheres Ziel zu erreichen. Im Einzelnen: In einem traditionellen, gewinnorientierten Unternehmen ist die Investitionsrendite tendenziell die wichtigste Anforderung jedes Businessplans und auch das Hauptziel oder der Plan des Planes. In den meisten traditionellen Unternehmen Rückkehr auf Investitionen neigt dazu, das Herz aller Aktivitäten sein, da in der Regel das Unternehmen existiert, um die Rendite (Gewinn und Wachstum effektiv) der Aktionäre Mittel in das Unternehmen investiert zu maximieren. Die Planung in traditionellen Unternehmen vergißt diese Grundverpflichtung, vor allem, wenn ein Junior Manager zum ersten Mal einen Businessplan schreiben will. In traditionellen gewinnorientierten Unternehmen, wenn ein neuer Manager beginnt, einen Business-Plan oder operativen Plan zum ersten Mal (und für einige erfahrene Manager auch zum zehnten Mal zu schreiben), fragt sich der Manager: Was ist das Ziel Was versuche ich Zu erreichen Oft, wenn sie ihre eigenen Manager fragen, hat der Manager die gleichen Zweifel. Das zentrale Ziel ist in der Regel die Rentabilität. In Unternehmen oder gemeinnützigen Organisationen, in denen die Anreize der Anteilseigner nicht der Hauptzweck sind, ist der Return on Investment weniger ein Treiber in der Unternehmensplanung, ist aber dennoch eine entscheidende Anforderung. Solche Unternehmen werden immer beliebter, und wird auch weiterhin so werden, da der Zusammenbruch der westlichen Volkswirtschaften im Jahr 2008 und zunehmende Enttäuschung mit alten Stil Denken. Hier Return on Investment ist nicht der primäre Treiber oder Ziel des Unternehmens. Stattdessen kann der Haupttreiber des Unternehmens ein anderer Zweck sein. Ein Beispiel für einen anderen Zweck könnte die Tätigkeit eines Sozialunternehmens oder einer Genossenschaft sein, oder vielleicht eine Arbeitnehmerbesitzfirma oder vielleicht ein Vertrauen oder eine Wohltätigkeitsorganisation, deren Hauptziel (eher als die herkömmliche Profitgenerierung für externe institutionelle Anteilseigner) sein kann Um lokale Arbeitsplätze zu sichern und um die lokale Gemeinschaft zu fördern oder um Wissenschaft oder Lernen oder Gesundheit voranzutreiben usw. Während die Rentabilität der Investitionen für die Planung und den Betrieb weniger entscheidend oder angemessen erscheint, muss das Unternehmen dennoch finanziell bleiben Lebensfähig. Oder er hört auf, überhaupt operieren zu können. In solchen Fällen wird die Investitionsrendite in die Unternehmensplanung in der Regel nicht maximiert, muss aber nach wie vor als Unterhaltsanforderung für Planung und Flexion nach den grundlegenden Zielen und finanziellen Anforderungen des Unternehmens behandelt werden. Vor der Planung ist es daher hilfreich, klar zu verstehen: Was streben wir eigentlich an Was ist unsere Poli - tik auf die soziale Verantwortung und Ethik der Gesellschaft usw. - unsere Philosophie Und was für ein Return on Investment (oder eine alternative finanzielle Performance) Ist dies ein strategischer Treiber für sich, oder einfach die Mittel, mit denen wir unsere Aktivitäten zur Unterstützung unserer (Punkt 1) Ziele planen - Ursache und Wirkung. Die grundlegende Methodik der Unternehmensplanung ist die Identifizierung von Ursachen und Wirkungen. Nach Ihren relevanten Unternehmensanforderungen (Finanzen und Ethik) und strategischen Fahrern (was wir eigentlich erreichen wollen). Hier ist eine Ursache eine Eingabe oder Aktion oder Ressource eine Wirkung ist ein Ergebnis oder Ergebnis oder Folge einer Art. Wir wollen einen xyz-Effekt erzielen (z. B. eine gegebene Rendite oder ein bestimmtes Umsatzniveau oder Marktanteil, egal was) - also was sollten wir planen, dies zu bewirken? Häufig werden große Kauseeffekt-Elemente in kleinere Aktivitäten unterteilt, die auch Umfassen eine Ursache und Wirkung. (Der Planungsprozess und die Hilfsmittel helfen, zu erklären, wie diese Unterteilung funktioniert - wo ein großes Ziel in kleinere, messbarere und erreichbare Teile zerlegt wird). Junior Manager haben Verantwortung für Pläne und Aktivitäten, die in größere Abteilungspläne und Aktivitäten von Senior Managern einfließen. Die Pläne und Aktivitäten von Führungskräften fließen in die Teilungspläne von Führungskräften und Direktoren ein. Es gibt eine Hierarchie oder Baumstruktur von Ursache und Wirkung, die alle hoffentlich zum gesamten organisatorischen Ziel beitragen. In vielen guten Geschäftsbereichen reicht eine wesentliche unternehmerische Planungsverantwortung für die Kunden vor Ort und die Tendenz steigt. In diesem Zusammenhang könnte der Businessplan auch als Marketing-Plan oder ein Vertriebs-Plan genannt werden - alle Abteilungspläne sind grundsätzlich Arten der Unternehmensplanung: Was werden Sie an wen verkaufen, wann und wie Sie es verkaufen werden , Wieviel Beitrag (Bruttogewinn) die Verkäufe produzieren wird, was die Marketing-andor Verkaufskosten sind und was die Rückkehr auf Investition sein wird. Wenn es sich bei einer Abteilung nicht um eine Kostenstelle handelt, sondern um ein Profit-Center, das Produkte oder Dienstleistungen intern an anderen Abteilungen anstatt extern anbietet, können sich die Sprach - und Planungselemente ändern, die Prinzipien bleiben jedoch dieselben. Diese Prinzipien und Methoden gelten auch für sehr große komplexe multinationale Organisationen, die mehr und mehr Kosten, feste Gemeinkosten, Einnahmen und damit größere Planungsformate beinhalten, mehr und größere Tabellenkalkulationen, mehr Zeilen und Spalten für jeden, mehr Aufmerksamkeit und Menschen arbeiten Auf die Zahlen, mehr Buchhalter, und typischerweise - vor allem auf der Ebene des mittleren Managements - mehr Wert auf Cashflow und Bilanz sowie auf die grundlegende Gewinn - und Verlustplanung. Führen Sie Ihre Marktforschung, einschließlich das Verständnis Ihrer Konkurrenz Aktivität Der Markt variiert je nach Unternehmen oder Organisation betroffen, aber jede organisierte Aktivität hat einen Markt. Die Kenntnis des Marktes ermöglicht es Ihnen, zu bewerten und zu bewerten und zu planen, wie man sich damit beschäftigt. Ein gemeinsames Misserfolg der Unternehmensplanung oder operativen Planung außerhalb der Geschäftswelt ist es, isoliert zu planen und nach innen zu schauen, wenn Ideen sehr positiv und zuverlässig scheinen können, weil es keinen Kontext und nichts zu vergleichen gibt. Daher ist die Forschung von entscheidender Bedeutung. Und das gilt für jede Art von Organisation - nicht nur für Unternehmen. Siehe insbesondere die Leitlinien für Marketing, da es sich auf die Unternehmensplanung bezieht. Planung sehr viel betrifft Prozesse. Die Grundsätze des Marketings wird auch erklären, wie man Bedeutung und Werte in das, was Sie planen setzen. Ihre Marktforschung sollte sich auf die Informationen konzentrieren, die Sie benötigen, um Ihnen zu helfen, Strategie zu formulieren und geschäftliche Entscheidungen zu treffen. Marktforschung sollte pragmatisch und zielgerichtet sein - ein Mittel zum Zweck und kein Mittel für sich. Marktinformationen decken ein breites Spektrum von Daten ab, von globalen Makro-Trends und Statistiken bis hin zu sehr spezifischen und detaillierten lokalen oder technischen Informationen, so dass es wichtig ist, zu entscheiden, was eigentlich relevant und notwendig zu wissen ist. Marktinformationen über Markt - und Branchentrends, Werte, Hauptkonzerne, Marktstrukturen usw. sind für große Konzerne, die auf nationaler oder internationaler Ebene tätig sind, wichtig. Diese Art von Forschung wird manchmal auch als sekundäre, weil es bereits vorhanden ist, wurde erforscht und veröffentlicht zuvor. Diese Informationen sind im Internet, Bibliotheken, Forschungsunternehmen, Handel und nationale Presse und Publikationen, Fachverbände und Institute erhältlich. Diese Sekundärforschung Informationen in der Regel erfordert einige Interpretation oder Manipulation für Ihre eigenen Zwecke. Jedoch theres kein Punkt, der Tageforschung globale statistische ökonomische und demographische Daten aufwendet, wenn Sie eine Strategie für ein verhältnismäßig kleines oder lokales Geschäft entwickeln. Viel nützlicher wäre es, Ihre eigene Primärforschung (d. H. Ursprüngliche Forschung) über den lokalen Zielmarkt durchzuführen, Muster und Präferenzen, lokale Konkurrenten, ihre Preise und Dienstleistungsangebote zu kaufen. Viele nützliche primäre Marktforschungen können mit Kunden-Feedback, Umfragen, Fragebögen und Fokusgruppen (Indikatoren und Ansichten durch Diskussion unter ein paar repräsentative Personen in einer kontrollierten Diskussion Situation) durchgeführt werden. Diese Art von Primärforschung sollte genau auf Ihre Bedürfnisse zugeschnitten werden. Primäre Forschung erfordert weniger Manipulation als sekundäre Forschung, aber alle Arten von Forschung müssen eine gewisse Menge an Analyse. Seien Sie vorsichtig, wenn Sie Zahlen extrapolieren oder projizieren, um zu vermeiden, dass Fehler oder falsche Annahmen vergrößert werden. If the starting point is inaccurate the resulting analysis will not be reliable. For businesses of any size small, local, global and everything in between, the main elements you need to understand and quantify are: customer (and potential customer) numbers, profile and mix customer perceptions, needs, preferences, buying patterns, and trends, by sub-sector if necessary products and services, mix, values and trends demographic issues and trends (especially if dependent on consumer markets) future regulatory and legal effects prices and values, and customer perceptions in these areas distribution and routes to market competitor activities, strengths, weaknesses, products, services, prices, sales methods, etc Primary research is recommended for local and niche services. Keep the subjects simple and the range narrow. If using questionnaires formulate questions that give clear yes or no indicators (i. e. avoid three and five options in multi-choices which produce lots of uncertain answers) always understand how you will analyse and measure the data produced. Try to convert data to numerical format and manipulate on a spreadsheet. Use focus groups for more detailed work. For large research projects consider using a market research organization because theyll probably do it better than you, even though this is likely to be more costly. If you use any sort of marketing agency ensure you issue a clear brief, and that your aims are clearly understood. Useful frameworks for research are PEST analysis and SWOT analysis. establish your corporate philosophy and the aims of your business or operation First establish or confirm the aims of the business, and if you are concerned with a part of a business, establish and validate the aims of your part of the business. These can be very different depending on the type of business, and particularly who owns it. Refer to and consider issues of ethics and philosophy, corporate social responsibility, sustainability. etc - these are the foundations on which values and missions are built. Consider the Psychological Contract and the benefits of establishing a natural balance and fairness between all interests (notably staff, customers, the organization). Traditional business models are not necessarily the best ones. The world is constantly changing, and establishing a new business is a good time to challenge preconceptions of fundamental business structure and purpose. A business based on a narrow aim of enriching a few investors while relegating the needs and involvement of everyone else may contain conflicts and tensions at a deep level. There are other innovative business structures which can inherently provide a more natural, cooperative and self-fuelling relationship - especially between employees and the organization, and potentially between customers and the organization too. When you have established or confirmed your philosophical and ethical position, state the objectives of the business unit you are planning to develop - your short, medium and long term aims - (typically short, medium and long equate to 1 year, 2-3 years and 3 years plus). In other words, what is the business aiming to do over the next one, three and five years Bear in mind that you must reliably ensure the success and viability of the business in the short term or the long term is merely an academic issue. Grand visions need solid foundations. All objectives and aims must be prioritised and as far as possible quantified. If you cant measure it, you cant manage it. define your mission statement All businesses need a 145mission statement. It announces clearly and succinctly to your staff, shareholders and customers what you are in business to do. Your mission statement may build upon a general 145service charter relevant to your industry. You can involve staff in defining and refining the businesss mission statement, which helps develop a sense of ownership and responsibility. Producing and announcing the mission statement is also an excellent process for focusing attention on the businesss priorities, and particularly the emphasis on customer service. Whole businesses need a mission statement - departments and smaller business units within a bigger business need them too. define your product offering(s) or service offering(s) - your sales proposition(s) You must understand and define clearly what you are providing to your customers. This description should normally go beyond your products or services, and critically must include the way you do business . and what business benefits your customers derive from your products and services, and from doing business with you. Develop offerings or propositions for each main area of your business activity - sometimes referred to as revenue streams, or business streams - andor for the sector(s) that you serve. Under normal circumstances competitive advantage is increased the more you can offer things that your competitors cannot. Good research will tell you where the opportunities are to increase your competitive advantage in areas that are of prime interest to your target markets. Develop your service offering to emphasise your strengths, which should normally relate to your business objectives, in turn being influenced by corporate aims and market research. The important process in developing a proposition is translating your view of these services into an offer that means something to your customer. The definition of your service offer must make sense to your customer in terms that are advantageous and beneficial to the customer, not what is technically good, or scientifically sound to you. Think about what your service, and the manner by which you deliver it, means to your customer. Traditionally, in sales and marketing, this perspective is referred to as translating features into benefits. The easiest way to translate a feature into a benefit is to add the prompt 145which means that. . For example, if a strong feature of a business is that it has 24-hour opening, this feature would translate into something like: Were open 24 hours (the feature) which means that you can get what you need when you need it - day or night. (the benefit). Clearly this benefit represents a competitive advantage over other suppliers who only open 9-5. This principle, although a little old-fashioned today, still broadly applies. The important thing is to understand your services and proposition in terms that your customer will recognise as being relevant and beneficial to them. Most businesses have a very poor understanding of what their customers value most in the relationship, so ensure you discover this in the research stage, and reflect it in your stated product or service proposition(s). Customers invariably value these benefits higher than all others: Making money Saving money Saving time If your proposition(s) cannot be seen as leading to any of the above then customers will not be very interested in you. A service-offer or proposition should be an encapsulation of what you do best, that you do better than your competitors (or that they dont do at all) something that fits with your business objectives, stated in terms that will make your customers think 145Yes, that means something to me and I think it could be good for my business (and therefore good for me also as a buyer or sponsor). This is the first brick in the wall in the process of business planning, sales planning, marketing planning, and thereafter, direct marketing, and particularly sales lead generation. write your business plan - include sales, costs of sales, gross margins, and if necessary your business overheads Business plans come in all shapes and sizes. Pragmatism is essential. Ensure your plan shows what your business needs it to show. Essentially your plan is a spreadsheet of numbers with supporting narrative, explaining how the numbers are to be achieved. A plan should show all the activities and resources in terms of revenues and costs, which together hopefully produce a profit at the end of the trading year. The level of detail and complexity depends on the size and part of the business that the plan concerns. Your business plan, which deals with all aspects of the resource and management of the business (or your part of the business), will include many decisions and factors fed in from the marketing process. It will state sales and profitability targets by activity. In a marketing plan there may also be references to image and reputation, and to public relations. All of these issues require thought and planning if they are to result in improvement, and particularly increasing numbers of customers and revenue growth. You would normally describe and provide financial justification for the means of achieving these things, together with customer satisfaction improvement. Above all a plan needs to be based on actions - cost-effective and profitable cause and effect inputs required to achieved required outputs, analysed, identified and quantified separately wherever necessary to be able to manage and measure the relevant activities and resources. quantify the business you seek from each of your market sectors, segments, products and customer groupings, and allocate investment, resources and activities accordingly These principles apply to a small local business, a department within a business, or a vast whole business. Before attending to the detail of how to achieve your marketing aims you need to quantify clearly what they are. What growth targets does the business have What customer losses are you projecting How many new customers do you need, by size and type, by product and service What sales volumes, revenues and contributions values do you need for each business or revenue stream from each sector What is your product mix, in terms of customer type, size, sector, volumes, values, contribution, and distribution channel or route to market What are your projected selling costs and net contributions per service, product, sector What trends and percentage increase in revenues and contributions, and volumes compared to last year are you projecting How is your market share per business stream and sector changing, and how does this compare with your overall business aims What are your fast-growth high-margin opportunities, and what are your mature and low-margin services how are you treating these different opportunities, and anything else in between You should use a basic spreadsheet tool to split your business according to the main activities and profit levers. See the simple salesbusiness planning tool example below. ansoff product-market growth matrix - strategic tool A useful planning tool in respect of markets and products is the matrix developed by Igor Ansoff (H Igor Ansoff, 1918-2002), who is regarded by some as the Father of Strategic Management. Fully titled the Ansoff Product-Market Growth Matrix, the tool was first published in Harvard Business Review, 1957, in Ansoffs paper Strategies for Diversification. The Ansoff product-market matrix helps to understand and assess marketing or business development strategy. Any business, or part of a business can choose which strategy to employ, or which mix of strategic options to use. This is a fundamentally simple and effective way of looking at strategic development options. Each of these strategic options holds different opportunities and downsides for different organizations, so what is right for one business wont necessarily be right for another. Think about what option offers the best potential for your own business and market. Think about the strengths of your business and what type of growth strategy your strengths will enable most naturally. Generally beware of diversification - this is, by its nature, unknown territory, and carries the highest risk of failure. Here are the Ansoff strategies in summary: market penetration - Developing your sales of existing products to your existing market(s). This is fine if there is plenty of market share to be had at the expense of your competitors, or if the market is growing fast and large enough for the growth you need. If you already have large market share you need to consider whether investing for further growth in this area would produce diminishing returns from your development activity. It could be that you will increase the profit from this activity more by reducing costs than by actively seeking more market share. Strong market share suggests there are likely to be better returns from extending the range of productsservices that you can offer to the market, as in the next option. product development - Developing or finding new products to take to your existing market(s). This is an attractive strategy if you have strong market share in a particular market. Such a strategy can be a suitable reason for acquiring another company or productservice capability provided it is relevant to your market and your distribution route. Developing new products does not mean that you have to do this yourself (which is normally very expensive and frequently results in simply re-inventing someone elses wheel) - often there are potential manufacturing partners out there who are looking for their own distribution partner with the sort of market presence that you already have. However if you already have good market share across a wide range of products for your market, this option may be one that produces diminishing returns on your growth investment and activities, and instead you may do better to seek to develop new markets, as in the next strategic option. market development - Developing new markets for your existing products. New markets can also mean new sub-sectors within your market - it helps to stay reasonably close to the markets you know and which know you. Moving into completely different markets, even if the productservice fit looks good, holds risks because this will be unknown territory for you, and almost certainly will involve working through new distribution channels, routes or partners. If you have good market share and good productservice range then moving into associated markets or segments is likely to be an attractive strategy. diversification - taking new products into new markets. This is high risk - not only do you not know the products, but neither do you know the new market(s), and again this strategic option is likely to entail working through new distribution channels and routes to market. This sort of activity should generally be regarded as additional and supplementary to the core business activity, and should be rolled out carefully through rigorous testing and piloting. Consider also your existing products and services themselves in terms of their market development opportunity and profit potential. Some will offer very high margins because they are relatively new, or specialised in some way, perhaps because of special USPs or distribution arrangements. Other products and services may be more mature, with little or no competitive advantage, in which case they will produce lower margins. The Boston Matrix is a useful way to understand and assess your different existing product and service opportunities: boston matrix model - productservice development The Boston Matrix model (also called the BSG Matrix, Growth-Share Matrix, and variations around these titles) is a tool for assessing existing and development products in terms of their market potential, and thereby implying strategic action for products and services in each of the four categories reflected in the model. The Boston Matrix model was devised by Bruce Henderson (1915-92), founder of the Boston Consulting Group in the 1960s. It has been adapted in many ways. A simple version is shown here below. Like other four-part 2x2 matrix models, the Boston Matrix is a very quick and easy method for analysis, thinking and decision-making, while being unavoidably limited in its handling of subtlety and detail. Often in business and strategic thinking too much detail is unhelpful - instead, clarity and ease of understanding are extremely helpful, especially in communicating ideas to teams and groups, in which circumstances the Boston Matrix is an excellent aid. low market share These simple split analysis tools are an extremely effective way to plan your sales and business. Construct a working spreadsheet so that the bottom-right cell shows the total sales or gross margin, or profit, whatever you need to measure, and by changing the figures within the split (altering the mix, average prices, quantities, etc) you can carry out what if analysis to develop the best plans. If you are a competent working with spreadsheets it is normally possible to assemble all of this data onto a single spreadsheet and then show different analyses by sorting and graphing according to different fields. When you are happy with the overall totals for the year, convert this into a phased monthly plan, with as many lines and columns as you need and are appropriate for the business. Develop this spreadsheet by showing inputs as well as sales outputs - the quantifiable activity (for example, the numbers of enquiries necessary to produce the planned sales levels) required to produce the planned performance. Large businesses need extensive and multiple page spreadsheets. A business plan needs costs as well as sales, and will show profit as well as revenue and gross margin, but the principle is the same: plan the detailed numbers and values of what the business performance will be, and what inputs are required to achieve it. Heres a free MSExcel profit and loss account template tool for incorporating these factors and financials into a more formal phased business trading plan, which also serves as a business forecasting and reporting tool too. Adapt it to suit your purposes. This plan example is also available as a PDF, see the Profit and Loss Account (PampL) Small Enterprise Business Plan Example (PDF). The numbers could be anything: ten times less, ten times more, a hundred times more - the principle is the same. Consider also indirect activities that affect sales and business levels, such as customer service. Identify key performance indicators here too, such as customer complaints response and resolution levels and timescales. Internal lead referral schemes, strategic partnership activity the performance of other direct sales activities such as sales agencies, distributorships, export activities, licensing, etc. These performance factors wont normally appear on a business plan spreadsheet, but a separate plan should be made for them, otherwise they wont happen. write your marketing plan or business plan Your marketing plan is actually a statement, supported by relevant financial data, of how you are going to develop your business. Plans should be based on actions, not masses of historical data. The historical and market information should be sufficient just to explain and justify the opportunities, direction, strategy, and most importantly, the marketing actions, methods and measures - not to tell the story of the past 20 years of your particular industry. What you are going to sell to whom, when and how you are going to sell it, how much contribution (gross profit) the sales produce, what the marketing cost will be, and what will be the return on investment. As stated above it is easiest and best to assemble all of this data onto a spreadsheet, which then allows data to be manipulated through the planning process, and then changed and re-projected when the trading year is under way. The spreadsheet then becomes the basis of your sales and marketing forecasting and results reporting tool. As well as sales and marketing data, in most types of businesses it is also useful to include measurable aims concerning customer service and satisfaction. The marketing plan will have costs that relate to a marketing budget in the overall business plan. The marketing plan will also have revenue and gross marginprofitability targets that relate to the turnover and profitability in the overall business plan. This data is essentially numerical, and so needs also some supporting narrative as to how the numbers will be achieved - the actions - but keep the narrative concise if it extends to more than a half-dozen sheets make sure you put a succinct executive summary on the front. The marketing plan narrative could if appropriate also refer to indirect activities such as product development, customer service, quality assurance, training etc. if significantly relevant to achieving the marketing plan aims. Be pragmatic - marketing plans vary enormously depending on the type, size and maturity of business. Above all create a plan that logically shows how the business can best consolidate and grow its successful profitable areas. The marketing plan should be a working and truly useful tool - if it is, then its probably a good one. sample business plan, marketing plan or sales plan sample structure and example formattemplate Keep the written part of the business plan as concise and brief as possible - most situations and high-ranking executives do not need to see plans that are an inch thick. If you can make your case on a half dozen pages then do so. Particularly if your plan is more than 5-6 pages long, produce an executive summary (easiest to do when you have completed the plan) and insert it at the beginning of the document. If you need to include lots of reference material, examples, charts, evidence, etc, show these as appendices at the back of the document and make sure they are numbered and referenced during the main body of the plan. Each new section should start at the top of a new page. Number the pages. Important plans should be suitably bound. All business plans should be professionally and neatly presented, with no grammar and spelling errors, clearly laid out in an easy to read format (avoid lots of upper-case or fancy fonts or italics as these are all difficult to read). Your business plan contents and structure should be as follows: business plans structure - a business planning template Title page: Title or heading of the plan and brief description if required, author, date, companyorganization if applicable, details of circulation and confidentiality. Contents page: A list of contents (basically the sections listed here, starting with the Introduction page) showing page numbers, plus a list of appendices or addendums (added reference material at the back of the document) allowing the reader to find what they need and navigate the document easily, and to refer others to particular items and page numbers when reviewing or querying. Introduction page . Introduction and purpose of the plan, terms of reference if applicable (usually for formal and large plans or projects). Executive summary page: Optional and usually beneficial, this should normally be no more than a page long (or its not an executive summary) - the key points of the whole plan including conclusions, recommendations, actions, financial returns on investment, etc. clearly readable in a few minutes. Main body of plan: sections and headings as required, see template below. Acknowledgments and bibliographyreference sources: if relevant (only required normally for very large formal plans) Appendices: appendices or addendums - additional detailed reference material, examples, statistics, spreadsheets, etc. for reference and not central to the main presentation of your plan. business plans - main body sections examples template This sample template is typical for a salesmarketingnew business development business plan. (A business plan for a more complex project such as an international joint-venture, or the formation of a new company including manufacturing plant or other overhead activities would need to include relevant information and financials about the overheads and resources concerned, and the financials would need to show costs and profits more like a fully developed profit and loss account, with cashflow projections, balance sheet, etc.) Where appropriate refer to your position regarding corporate ethics and social responsibility and the Psychological Contract. While these aspects are not mechanisms within the plan, they are crucial reference points. Define your market - sector(s) and segment(s) definitions Quantify your market (overview only) - size, segmentation, relevant statistics, values, numbers (locations, peopleusers, etc) - make this relevant to you business Explain your market(s) - sector trends, eg. growth, legislation, seasonality, PEST factors where relevant, refer to Ansoff matrix, show the strategic business drivers within sector and segments, purchasing mechanisms, processes, restrictions - what are the factors that determine customers priorities and needs - this is a logical place to refer to ethics and CSR (corporate social responsibility Explain your existing business - your current business according to sector, productsservices, quantities, values, distributor, etc. Analyse your existing customer spread by customer type, values and productsservices including major accounts (the Pareto Principle or the 80:20 rule often applies here, eg. 80 of your business comes from 20 of your customers) Explain your products and services - refer to Boston matrix and especially your strategic propositions (what these propositions will do for your customers) including your USPs and UPBs (see sales training section and acronyms ) Explain you routes to market, gatekeepers, influencers and strategic partners - the other organizationsindividuals you will work with to develop your market, including whats in it for them, commissions, endorsements, accreditations, approvals, licenses, etc. Case studies and track record - the credibility, evidence and proof that your propositions and strategic partnerships work Competitor analysis, eg. SWOT analysis of your own business compared to SWOT analysis of each competitor Salesmarketingbusiness plan (1 year min) showing sales and margins by productservice stream, mix, values, segment, distributor, etc, whatever is relevant, phased monthly, in as much detail as you need. This should be on a spreadsheet . with as many different sheets as necessary to quantify relevant inputs and outputs. List your strategic actions (marketing campaigns, sales activities, advertising, etc) that will deliver the above, with costs and returns. This should be supported with a spreadsheet, showing cost and return on investment for each activity. Tip: If the business plan concerns an existing activity, use the previous years salesbusiness analysis as the basis for the next years salesbusiness plan. Adapt as necessary according to your new strategic plans. other business planning and marketing issues staffing and training implications Your people are unlikely to have all the skills they need to help you implement a marketing plan. You may not have all the people that you need so you have to consider justifying and obtaining extra. Customer service is acutely sensitive to staffing and training. Are all of your people aware of the aims of the business, its mission statement and your sales propositions Do they know what their responsibilities are How will you measure their performance Many of these issues feed back into the business plan under human resources and training, where budgets need to be available to support the investment in these areas. customer service charter You should formulate a customer service charter, extending both your mission statement and your service offer, so as to inform staff and customers what your standards are. These standards can cover quite detailed aspects of your service, such as how many times the telephone will be permitted to ring until the caller is gets an answer. Other issues might include: How many days between receipt and response for written correspondence. Complaints procedure and timescales for each stage. This charter sets customer expectations, so be sure you can meet them. Customers get disappointed particularly when their expectations are not met, and when so many standards can be set at arbitrary levels, think of each one as a promise that you should keep. Business-to-business customers would expect to agree these standards with their suppliers and have them recorded as part of their contracts, or as SLAs (service level agreements). Increasingly, large customers demand SLAs to be tailored to their own specific needs, and the process of developing these understandings and agreements is absolutely crucial to the maintenance and development of large contracts. Remember an important rule about customer service: Its not so much the failure to meet standards that causes major dissatisfaction among customers - everyone can make a mistake - the biggest cause of upset is the failure of suppliers to inform customers and keep them updated when problems arise. Not being told in advance, not receiving any apology, not getting any explanation why, and not hearing whats going to be done to put things right, are key areas of customer dissatisfaction, and therefore easy areas for suppliers to focus their efforts to achieve and communicate improvements. A special point of note for businesses that require a strong technical profile among their service staff: these people are often reactive by nature and so not good at taking initiative to identify and anticipate problem areas in customer service. Its therefore helpful to establish suitable mechanisms and responsibility to pick up problems and deal with them - a kind of trouble-shooting capability - which can be separately managed and monitored at a strategic level. Do not assume that technically-oriented staff will be capable of proactively developing customer service solutions and revisions to SLAs - they generally need help in doing so from staff with high creativity, empathy, communications and initiative capabilities. establish systems to measure customer service and staff performance These standards and the SLAs established for large customers need to be visible, agreed with customers, absolutely measurable. You must keep measuring your performance against them, and preferably publishing the results, internally and externally. Customer complaints handling is a key element: Measuring customer complaints is crucial because individual complaints are crucial areas to resolve, and also as a whole, complaints serve as a barometer for the quality and performance of the business. You need to have a scheme which encourages, not discourages, customers to complain, to open the channels as wide as possible. Most businesses are too defensive where complaints are concerned, preferring to minimise their importance, or to seek to justify and excuse them. Wrong. Complaints are the opportunities to turn ordinary service into unbeatable service. Moreover, time and again surveys suggest that anything up to nine out of ten people do not complain to the provider when they feel dissatisfied - they just keep their dissatisfaction to themselves and the provider never finds out theres a problem, even when the customer chooses to go elsewhere. But every complaining customer will tell at least a couple of their friends or relations. Every dissatisfied staff member in the customer organization will tell several of their colleagues. Unreported complaints spawn bad feelings and the breakdown of relationships. It is imperative that you capture all complaints in order to: Put at ease and give explanation or reassurance to the person complaining. Reduce the chances of them complaining to someone else. Monitor exactly how many dissatisfied customers you have and what the causes are, and thats even more important if youre failing to deliver your mission statement or service offer Take appropriate corrective action to prevent a re-occurrence. If appropriate (ie for large customers) review SLAs and take the opportunity to agree new SLAs with the customer. implications for IT, premises, and reporting systems Also relating to your business plan are the issues of: Information Technology - are your computers and communications systems capable of giving you the information and analysis you need How do you use email - is it helping or hindering your business and the quality of service you give to your customers What internet presence and processes do you need How should your voice and data systems work together What systems need to be available to mobile staff What customer relationship management (CRM) systems should you have How should you consider all these issues to see the needs and opportunities IT and communications systems increasingly offer marketing and competitive advantage to businesses in all sectors - make sure you know hat IT can do for you and for your customers. Premises - Review your premises and sites in light of your customer service, distribution, and customer relationship requirements. Pay particular attention anywhere in your organization that your customers visit - the impression and service you give here is critical. Reporting systems - If you cant measure it you cant manage it, and where finance and business performance is concerned this is certainly true. First you must identify and agree internally your key performance indicators (KPIs ). Identify every aspect of your service or performance that is important - then you need to be able to measure it and report on it, and where people are involved in performing to certain standards then the standards and the reporting needs to be transparent to them also. How do you report on sales, marketing and business performance and interpret the results Who needs to know Who needs to capture the data communications and ongoing customer feedback are essential Having an open dialogue with your customers is vital. Theres a double benefit to your business in ensuring this happens: You nip problems in the bud and stay aware of how youre performing. Your customers feel better about the service you provide as a result of the communications, or from the fact that the channel is open even if they dont use it - its human nature. Try to devise a standard feedback form. It can double as a promotional tool as well if its made available on a wider scale. The form can carry details of your mission statement, service offer and your customer service charter. Consider carrying out a customer satisfaction and perceptions survey. There are many ways to do this on a small or large scale, and valuable feedback is always obtained from customer survey exercises. tips for starting a small business or self-employment - for non-financial people Some of us are not naturally inclined towards the sort of detailed financial thinking that is required for traditional detailed business planning. If this is you, youll possess other valuable capabilities that will be useful in your own enterprise, and youll maybe find it helpful to use this alternative approach to planning a new enterprise or self-employment. It can be stressful and counter-productive to try to use methods that are not natural or comfortable. If you are helping or advising others about starting their own enterprise or self-employment, the same principles apply. Not everyone is naturally good at business planning, but everyone who dreams of being self-employed or who wants to start and run their own independent enterprise is capable of doing so, provided they work to their strengths, capabilities and passions. People running successful enterprises come in all shapes and sizes, from all backgrounds, all ages, with skills, passions, and capabilities in any field you can imagine. Anyone can run their own business or be successful in self-employment given the simple determination to do so. Business and enterprise is not just for stereotypical business-types the benefits and advantages of being your own boss are available to us all. Here are some pointers for people considering starting their own new enterprise, or for helping others to do the same. First, and especially if you are not clear of your own real strengths, or what direction to pursue, focus on using tools to understanding your own personality style and strengths. Then use this knowledge to imagine and realise how your natural capabilities can be used to best effect in defining and providing your own services or running your own enterprise. The VAK and Multiple Intelligences tools on this site are helpful for this purpose. They assess peoples strengths completely differently to traditional IQ or academic evaluations, which are extremely narrow and generally not relevant at all for people who want to be their own boss. Understanding personality is also useful since personality-type greatly influences the way that a person approaches self-employment or running an enterprise, and what sort of service or business to offer. The Personality Styles page provides a lot of explanation about this. Many people are conditioned by schools and over-cautious parents to under-estimate their own potential and capabilities, which is a big reason to take a fresh look at what you are good at, and to re-think and understand better the ways that your personality type tends to be successful in life and business. There are many ways to be successful and independent in life aside from building and running a conventional business and adhering to conventional financial planning methods. The basic economics of becoming successfully independent in any sort of venture are actually extremely simple, and focusing on the following simple fundamentals (a process really) can help many folk turn your dream or an idea into a successful enterprise or self-employment reality. Its usually easiest to think first of these factors in terms of daily, weekly or monthly numbers and values, and then to extend the figures to give totals for a whole year: 1. Whats your product or service (Whats goodspecialdifferent about your products or service that enough people will buy it And importantly is this something that you have a real passion for All successful enterprises are built on doing something the owner enjoys.) 2. What does it cost to makebuy inprovide the product or service (If you are buying and selling products or using materials consider the cost prices. If the main resource is your own time then attach a cost to your labour that reflects your available time for the work and the wage you need to draw. Divide your required annual wage by the number of work hours available to you, and this is your notional hourly labour cost.) 3. What price will the productservice sell for (Ideally small businesses need a healthy profit margin or mark-up - doubling the cost is good if the market will accept it. A mark-up of less than 50 is cause for concern unless you are selling products in relatively high volumes or values. Price your productsservices according to what the market will pay, not according to your costs. Take into account your competitors and what they charge and their relative quality. Service businesses that use only the persons time are often very attractive and profitable because there is no added complication of buying and holding stock - hence why window-cleaning, sign-writing, repairs, gardening, decorating, tutoring, writing, therapy, training, coaching and consultancy, etc. are such good businesses for people who prefer a simple approach to self-employment and enterprise. Consider the effect of VAT especially for consumer businesses - ie. selling to the general public - assuming your business is or must be VAT registered. Private consumers of course are more sensitive to VAT than business customers who can generally reclaim VAT should you have to add it to your prices.) 4. Who will buy the productservice (Identify your customers and market. Do you know this for sure Test your assumptions: this is a critical part of the proposition and generally benefits from more thought and research to confirm that a big enough market exists for your idea. Consider your competition - what are people buying currently and why will they buy from you instead) 5. How muchmany do you need to sell in a year And how many customers do you need (This is a vital part of the proposition to confirm that the gross profit (the difference between costs of bought in productslabour and sales revenues) covers yourtheir financial needs (including a living wage and other fixed costs of running the enterprise. Again remember the affect of VAT on your selling prices if applicable.) 6. How will people know about the serviceproduct (You need to understand what advertisingmarketingenquiry-generation is necessary - activity and cost. There is usually a cost for generating new customers, especially in the early stages of a new enterprise. Once the business is established, say after six months to a year, word-of-mouth referrals are for some businesses all that is required to produce new customers - especially those based in a local community, but virtually any new enterprise requires marketing at its launch. See the articles on marketing and selling .) 7. Does all this add up, and better still provide a cash surplus at the end of a year - if so then its probably a good business model. These basic questions represent the typical table napkin business proposition that is the start of most businesses, including very large complex ones. People who dislike and are not fluent in detailed business calculations might find the above process a useful starting point when thinking about how to begin a new enterprise or a venture in self-employment. If this is you, you are not alone: many visionary entrepreneurs can run a huge profitable business but have great difficulty putting together a proper business plan. Hence many highly successful business leaders rely heavily on their financial directors to take care of the financial details, leaving them free to get on with the business activity that makes best use of their natural skill, be it creativity, selling, service-provision, people-skills, technical skills, or whatever. Incidentally the above factors are the essential components which make up a basic Profit and Loss Account, which is the primary management tool for a business of any scale and complexity. Heres a free MSExcel profit and loss account template tool for extending these factors and financials into a more formal phased plan, which also serves as a business forecasting and reporting tool too. If in doubt about this seek some help from an experienced business person or your accountant. Adapt it to suit your purposes. The example PampL trading plan is also available as a pdf. The numbers could be anything - ten times less, ten times more, a hundred times more - the principle is the same. company types and financial set up - quick guide When you have confirmed and refined the basic viability of your business idea you can then begin getting to grips with the more detailed aspects of forming the business itself. This necessarily includes deciding your type of business constitution - the legal format of your company - or company type as it is often described. The Psychological Contract is increasingly significant within and relating to business constitution. Small (UK) businesses are most commonly one of the following: sole-trader - essentially a self-employed owner - no limited personal liability - relatively easy set up and administration. partnership - essentially a group of self-employed partnersowners - no limited personal liability - easy-ish set up and administration, although ultimately dependent on the complexity of the company and partnership. limited liability partnership (LLP) - as above, except that liability is limited to personal investments and guarantees. limited company (abbreviated to Ltd after the company name) - liability is limited to the assets of the company - registered with Companies House and legally obliged to publish accounts. There are less common variations of limited companies, and other business structures and constitutions, for example: social enterprise - various structures including. trusts, associations and especially cooperatives - these are not common typical or traditional business structures, but social enterprises are growing in popularity, and will be explained in more detail on this website in due course. Meanwhile here is useful information about cooperatives. public limited company (plc) - not appropriate for small companies. Sole-trader and partnership companies are very easy to set up and administer, but the ownerpartners are personally liable for all business debts and potential claims, so good insurance cover (including professional indemnity and public liability) is essential especially if business liabilities are potentially serious. A limited liability partnership offers protection to partners in terms of personal liabilities, in that liabilities are limited to the extent of personal investment and any other guarantees. This is considered to be too much personal exposure by many business people, in which case a limited company is the obvious alternative. A limited company exists in its own right - a tricky concept to understand for many people - basically meaning that financial liabilities belong to the company (its shareholders, to the value of their shares in other words) rather than the directors and executives of the business, as would apply in a partnership. Limited companies ultimately offer more flexibility for large complex businesses but can be over-complicated and administratively heavy if all you want to do is run a local shop or landscape gardening business or modest training or coaching business. Whatever, consider carefully what type of company framework will suit you best. Once established it can be quite difficult to unravel and change if you get it wrong - not impossible, but a nuisance if you could have got it right first time with a bit of extra thought at the planning stage. A good accountant will help you decide what is best for your situation from a legal and financial standpoint, although before this you should think for yourself what sort of business structure best fits your wider business situation, and especially your business aims and philosophy. Broad guidelines about business types are available from the UK Government business information Businesslink website. Youll need a business bank account. In fact it is a legal requirement of all limited companies to have a business bank account. Shop around. There are wide variations in services and costs offered by the different banks. You must also understand and organize the tax implications for your type of business. Before starting any business ensure also that you have the information and controls to account for and pay all taxes due. Helpfully to learn more about this in the UK, most tax affairs are within the responsibilities of HM Revenue and Customs - until they too change their name to something very silly. That said, the relevance today of HM (Her Majestys) is a bit puzzling when you stop to think about it and surely due for updating to the modern age. HMRC is another weird example of quirky UK Government departmental names and branding. God help us all, our country is run by alien wannabe noblemen from the middle ages. VAT (Value Added Tax or your national equivalent) is an issue warranting serious thought if your business is small enough to have a choice in the matter. Beyond a certain turnover (pound68,000 as at 2010) any UK business must register for VAT. Check the HMRC website for the current position. Being VAT registered means you must charge VAT on all VAT-rated supplies, which means also that the VAT you receive on payments from your customers must be paid to HM Revenue and Customs. (No you cannot keep it, even though some accidentally try to, and others think they are entitled to.) Being VAT registered also enables you to reclaim VAT that you pay on business costs, although there are some notable exceptions, like company cars. Retail and consumer businesses are especially affected by VAT. Private consumers cannot claim back VAT, so the effect of VAT on pricing and margins needs careful thought in planning any consumer business. Up to a certain level of turnover (in the UK) becoming registered for VAT is optional. If your business turnover is likely to be below the threshold for mandatory VAT registration, you must decide for yourself if the advantages outweigh the disadvantages. The main advantages of VAT registration are: your business will be perceived by certain people - especially other businesses - to be larger and more credible (not being registered for VAT indicates immediately that your turnover is below the VAT threshold) you will be able to reclaim VAT that you are charged on legitimate allowable business costs The main disadvantages of being VAT registered are: the administrative burden in keeping VAT records and submitting VAT returns (although this has been enormously simplified in recent years so that for small simple businesses it is really not a problem at all) risks of getting onto cashflow difficulties if you fail to set funds aside to pay your VAT bills (see the tax tips below) Information about VAT (and all other tax issues) is at the UK Government HM Revenue and Customs website: hmrc. gov. uk VAT is not the only tax. Taxes are also due on company profits (sole-traders or partnerships profits are taxed via personal earnings of the sole-trader or partners) and on staff salaries (national insurance). A sole-trader or partnership can employ staff, in which case national insurance tax is due on salaries paid to employees, which is different to the tax that employees pay themselves. Failing to retain funds in a company to pay taxes is a serious problem thats easily avoided with good early planning. Contact your tax office. Inform them of your plans and seek their help. Tax offices are generally extremely helpful, so ask. You can even talk to a real person on the phone without having to breach a six-level automated menu system. Ideally find a decent accountant too. Preferably one who comes recommended to you. With all the greatest respect to accountants everywhere, accountants are quite commonly very intense people, like solicitors and scientists, very much focused on process, accuracy, rules, etc. which in terms of personality fit can be a little at odds with the style of many entrepreneurs. So again shop around and find an accountant with whom you can share a joke and a beer or something from the human world. The relationship between a business person and hisher accountant is crucial if the business is to grow and develop significantly. Accountants might seem at times to be from another planet, but I can assure you the good ones are bloody magicians when it comes to business development, especially when the figures get really interesting. The statement that one stroke of an accountants pen is mightier than the worlds most successful sales team, is actually true. For many entrepreneurs, the ideal scenario is to grow your business large enough to support the cost of a really excellent finance director, who can take care of all the detailed legal and financial matters for you, and leave you completely free to concentrate on growing the business - concentrating your efforts and ideas and strategy externally towards markets and customers, and internally towards optimizing innovation and your staff. See the quick tax tips below, especially for small businesses which might not easily be able to achieve immediate and accurate control of their tax liabilities, which is one of the major early risks for a new successful small business. tax tips - understanding and accounting for taxes from the start A significant potential problem area for newly self-employed people, and for new business start-ups, is failing to budget and save for inevitable taxes which arise from your business activities. N. B. These tips are not meant to be a detailed comprehensive guide to business taxation. This section merely addresses a particular vulnerability of new start-up businesses in failing to set aside sufficient reserves to meet tax liabilities, especially small businesses, and even more especially sole-traders and partnerships and small limited companies, which lack expertise in accounting and consequently might benefit from these simple warnings and tips related to tax liabilities. In general these issues would normally be managed via a cashflow forecast, together with suitable financial processes to allocate and make payments for all costs and liabilities arising in the course of trading. I recognise however that many small business start-ups do not begin with such attention to financial processes, and its primarily for those situations that these particular notes are provided. These notes in no way suggest that this is the normal fully controlled approach to planning and organizing tax liabilities and other cashflow issues within any business of significant scale. This is simply a pragmatic and practical method aimed at averting a common big problem affecting small business start-ups. While your type of company and business determines precisely which taxes apply to you, broadly taxes are due on sales (for VAT registered businesses in the UK, or your VAT equivalent if outside the UK), and on the profits of your business and your earnings. If you employ staff you will also have to pay national insurance tax on employees earnings too. Generally sole-traders and partnerships have simpler tax arrangements - for example, profits are typically taxed as personal earnings - as compared with the more complex taxes applicable to limited companies, which also pay taxes on company profits and staff salaries. Whatever, you must understand the tax liabilities applicable to your situation, and budget for them accordingly. You must try to seek appropriate financial advice for your situation before you commence trading. Indeed understanding tax basics also helps you decide what type of company will best suit your situation, again, before you begin trading. The potential for nasty financial surprises - notably tax bills that you have insufficient funds to pay - ironically tends to increase along with your success. This is because bigger sales and profits and earnings inevitably produce bigger tax bills (percentage of tax increases too in the early growth of a business), all of which becomes a very big problem if youve no funds to pay taxes when due. The risks of getting into difficulties can be greater for the self-employed and small partnerships which perhaps do not have great financial knowledge and experience, than for larger Limited Company start-ups which tend to have more systems and support in financial areas. Start-ups are especially prone to tax surprises because the first set of tax bills can commonly be delayed, and if you fail to account properly for all taxes due then obviously you increase the chances of spending more than you should do, resulting in not having adequate funds to cover the payments when they are due. Risks are increased further if you are new to self-employment, previously having been employed and accustomed to receiving a regular salary on which all taxes have already been deducted, in other words net of tax. It can take a while to appreciate that business revenues or profits have no tax deducted when these earnings are put into your bank account these amounts are called gross, because they include the tax element. Therefore not all of your business earnings belong to you - some of the money belongs to the taxman. Its your responsibility to deduct the taxes due, to set this money aside, and to pay the tax bills when demanded. Additionally, if you are a person who is in the habit of spending everything that you earn, you must be even more careful, since this tendency will increase the risks of your being unable to pay your taxes. Failing to get on top of the reality of taxes from the very beginning can lead to serious debt and cashflow problems, which is a miserable way to run a business. So you must anticipate and set aside funds necessary to meet your tax liabilities from the very start of your business, even if you do not initially have a very accurate idea of what taxes will be due, or you lack effective systems to calculate them - many small start-ups are in this position. Nevertheless it is too late to start thinking about tax when the first demands fall due. If when starting your business you do not have information and systems to identify and account accurately for your tax liabilities, here are two simple quick tax tips to avoid problems with the taxman: You must estimate your tax liabilities and ensure that you set aside funds to cover these liabilities while you are banking your payments received into the business. The easiest way to do this is to identify the taxes applicable to your business, for example VAT and your own personal income tax and national insurance. Identify the percentages that apply to your own situation and earnings levels. You can do this approximately. It does not need to be very precise. Add these percentages together, and then set aside this percentage of all your earnings that you receive into your business. Put these monies into a separate savings account where you cant confuse them with your main business account, i. e. your working capital typically held in a current account. Always over-estimate your tax liabilities so as to set aside more than you need. Having a surplus is not a problem. Having not enough money to pay taxes because youve under-estimated tax due is a problem sometimes enough to kill an otherwise promising business. Heres an example to show how quickly and easily you can plan and set aside a contingency to pay your tax bills, even if youve no experience or systems to calculate them precisely. This example is based on a self-employed consultancy-type business, like a training or coaching business, in which there are no significant costs of sales (products or services bought in) or overheads, i. e. revenues are effectively the profits too, since there are minimal costs to offset against profits: example of estimating and setting aside money to pay taxes 1. In the UK VAT on most products and services is 17.5. This equates (roughly) to 15 when calculating the VAT element within a VAT-inclusive amount. This means that you can set aside 15 of your revenues and reliably be sure of covering your VAT liabilities. 2. In the UK personal income tax and national insurance combined is roughly 30 of earnings up to about pound30,000 (a little over in fact), rising to 49 - call it 50 - of earnings above pound30k - roughly. N. B. Income tax and national insurance are calculated on taxable earnings, which exclude money spent on legitimate business costs, and VAT received. These figures in the above example are approximate I emphasise again, which is all you need for this purpose, moreover the approximations are on the high side of what the precise liabilities actually are. Accountants call this sort of thinking prudent. Its a pessimistic approach to forecasting liabilities rather than optimistic, which is fundamental to good financial planning and management: if the pessimism is wrong then you end up with a surplus (which is good), but if you are wrong in making optimistic forecasts and estimates (over-ambitious sales, and lower-than-actual costs and liabilities), then you run out of money (which is bad). Back to the percentages. Knowing the income tax percentages enables you to set aside a suitable percentage of your earnings when you receive them into the business. Roughly speaking, for earnings up to pound30k you need to set aside 30 to cover income tax and national insurance. For earnings over pound30k you need to set aside 50 to cover your income tax and national insurance. (Earnings below pound30k remain taxable at 30). Remember you can arrive at these figures based on the VAT exclusive revenues, but to keep matters simpler it is easier to use an adjusted total percentage figure to apply to the total gross earnings. If its kept very simple and quick youll be more likely to do it - andor to communicate the method effectively to your partner if they are responsible for handling the financials, as often happens. Given this example, if in your first year your gross revenues (banked payments received) are say pound50,000, assuming you are VAT registered, then your tax liabilities will be (roughly): 17.5 VAT liabilities equates to 15 of gross sales revenues (again we are assuming no significant costs to offset these figures) (pound22.75k total tax divide pound50k gross revenues 45.5) From this example you can see that setting aside 45.5 of earnings (yes its a lot isnt it - which is why you need to anticipate it and set the money aside) would comfortably cover VAT and income tax liabilities. To be extra safe and simpler in this example you could round it up to 50. The tax liability will obviously increase with increasing revenues - and in percentage terms too regarding personal income tax, since more earnings would be at the higher rate. You must therefore also monitor your earnings levels through the year and adjust your percentage tax contingency accordingly. As stated already above, the risk of under-estimating tax liabilities increases the more successful you are, because tax bills get bigger. In truth you will have some costs to offset against the earnings figures above, but again for the purposes of establishing a very quick principle of saving a fixed percentage as a tax reserve until you know and can control these liabilities more accurately, the above is a very useful simple easy method of initially staying solvent and on top of your tax affairs, which are for many people the most serious source of nasty financial surprises in successful start-up businesses. The above example is very simple, and is provided mainly for small start-up businesses which might otherwise neglect to provide for tax liabilities. The figures and percentages are not appropriate (but the broad principle of forecasting and providing funds for tax liabilities is) to apply to retail businesses for example, or businesses in which staff are employed, since these businesses carry significant costs of sales and overheads, which should be deducted from revenues before calculating profits and taxes liabilities. Neither does the example take account of the various ways to reduce tax liabilities by reinvesting profits in the business, writing off stock, putting money into pensions, charitable donations, etc. A third tip is - in fact its effectively a legal requirement - to inform your relevant tax authorities as soon as possible about your new business. Preferably do this a few weeks before you actually begin trading. That way you can be fully informed of the tax situation - and your best methods of dealing with tax, because there are usually different ways, and sometimes the differences can be worth quite a lot of money. I do not go into more detail about tax here because its a very complex subject with wide variations depending on your own situation, for which you should seek relevant information and advice from a qualified accountant andor the relevant tax authorities. template and structure for a feasibility study or project justification report First, and importantly, you need to clarifyconfirm the criteria that need to be fulfilled in order to justify starting or continuing the project or group, in other words, what do the decision-makers need to see in order to approve the project or its continuation . Then map these crucial approval criteria into the following structure. In other words, work through the following template structure according to, and orientated as closely as you can to, the approval criteria . (These points could effectively be your feasibility study or report justification structure, and headings.) past, present and particularly future (customer) need (for the outputsresults produced by group or project) benefits and outcomes achieved to date for what costinvestment benefits and outcomes to be produced in the future resources, costs, investment . etc. required to produce future required outcomes and benefits (identify capital vs revenue costs, i. e. acquisition of major assets and ongoing overheads) alternative methods or ways of satisfying needs, with relative costreturn (return on investment) comparisons (ie. what other ways might there be for satisfying the need if the group or project doesnt happen or ceases) outline strategy and financial plan . including people, aims, philosophy . etc (ideally tuned to meet the authorising powers fulfilment criteria) for proposed start or continuation of project (assuming you have a case, and assuming there is no better alternative) Keep it simple. Keep to the facts and figures. Provide evidence. Be clear and concise. Refer to the tips about effective writing. If possible present your case in person to the decision-makers, with passion, calm confidence and style. Look at the tips on presentations. and assertiveness. tips on finding and working with business planning advisors and consultants If you need help putting together a business plan, and if you want to get the best from the engagement, its important to find the right person to work with, and to establish and maintain a good working relationship with them. If you are great big organisation youll probably not need to work with outsiders, and if you do then youll probably opt for a great big supplier, however there are significant benefits from working with much smaller suppliers - even single operators - and if you are a small business yourself, then this is probably the best choice anyway: to seek a good single operator, or small partnership of experts. Here are some ideas of what to look for. Youll be best finding someone who meets as much of this criteria as possible: lives close-by you so you can work face-to-face with them and get to know each other properly, and so that their time is efficiently used, instead of being in traffic on their way to and from your place is high integrity and very discreet is grown-up and got no baggage or emotional triggers - wise and mature - and it neednt be an age thing can help you see and decide where and how you want to take the business, rather than tell you where heshe thinks you need to go - a mentor not an instructor understands or can immediately relate to your industry sector and type of work is experienced working with small family companies, but is also a big picture strategist and visionary (advisors whove only ever worked with big corporations can sometimes be a bit free and easy with relatively small amounts of money - you need someone with a very very practical approach to managing cash-flow, and real business realities, whove worked in situations without the protection of vast corporate bureaucracy and the lack of transparency that this often brings) is triple-brained or whole-brained - mostly front-brained - (see the stuff on Benziger ) - intuitive-creative, thinking, but also able to be personable and grounded, subject to the point below complements your own strengths and fills the gaps and weaknesses in your collective abilities (again see the stuff on Benziger and Jung etc) - ie. if collectively you need hard facts and figures and logic then seek people with these strengths - conversely if you are strong on all this, then seek the creative humanist ethical strengths - heshe must work with you in a balanced team - so that the team has no blind spots, and no subjective biases in style or emphasis has two or three referees you can talk to and see evidence of past work (although if you check most of the above it will be a formality) doesnt smoke or drink too much isnt desperate for the work As regards finding someone like this, without doubt the most reliable and quickest method is by networking introductions through trusted people. The person you seek might be three or more links away, but if its a friend or associate of someone trusted, by someone whos trusted, by someone you trust, then probably theyll be right for you. Start by talking to people you know and asking if they know anyone, or if they know anyone who might know anyone - and take it from there. The chances of finding the right person in the local business listings or directory, out of the blue and from cold, are pretty remote. Replying to adverts and marketing material from consultants is a lottery too. Youll find someone eventually but youll need to kiss a lot of frogs first, which takes ages and is not the cleverest way to spend your valuable time. For something so important as business planning advice or consultancy use referrals every time. Referrals work not only because you get to find someone trusted, but the person you find has a reasonable assurance that you can be trusted too, you see: good suppliers are just as choosy as good clients. It works both ways. Be prepared to reward the person in whatever way is appropriate and fair (Im thinking percentage share of incremental success beyond expectations - perhaps even equity share if the person is really good and youd value their on-going contribution and help). Often the best people wont ask for much money up front at all, but from your point of view you will attract a lot more commitment and work beyond the call of normal duty from them if you reward higher than they ask or need. Good suppliers are immensely motivated by good clients and lots of appreciation, even if they dont want the financial reward. Good suppliers have usually seen too many ungrateful greedy people taking them for granted and penny pinching, and will tend to sack clients like these without even telling them why, and move on to more deserving enjoyable work with people who are fair and appreciative, which is how youll be Im sure. Finally, when youve found the right person, always continually agree expectations and invite feedback about how the relationship is working, not just how the work is going. starting your own business - or starting any new business These are the simple rules for planning and starting your own business. The principles also apply to planning and starting a new business within an organisation for someone else. In amongst the distractions and details of new business planning, it is important to keep sight of the basic rules of new business success: Your successful new business must offer something unique that people want. Uniqueness is vital because otherwise there is no reason for customers to buy from you. Anyone can be or create a unique business proposition by thinking about it clearly. Uniqueness comes in all shapes and sizes - its chiefly being especially good and different in a particular area, or field or sector. Uniqueness can be in a product or service, or in a trading method, or in you yourself, or any other aspect of your business which makes what you are offering special and appealing to people. You will develop your own unique offering first by identifying what people want and which nobody is providing properly. Second you must ensure that your chosen unique offering is also an extension of your own passion or particular expertise or strength - something you will love and enjoy being the best at - whatever it is. Every successful business is built on someones passion. new business start-ups by older people If you already have a career behind you, and you wonder if youve got it in you to compete and succeed in the modern world, consider this. First - you have definitely got it in you to succeed. Experience and wisdom are fundamental building blocks of success, and will be for you from the moment you start looking at yourself in this way. The reassuring wisdom that older people generally possess is extremely helpful in forming trusting relationships - with customers, suppliers, partners, colleagues, etc - which are essential for good business. Added to this, as we get older we have a greater understanding of our true passions and capabilities we know our strengths and styles and tolerances. This gives older people a very special potency in business. Older people know what they are good at. They play to their strengths. They know which battles they can win, and which to avoid. Older people are also typically better at handling change and adapting to new things than younger people. This is because older people have had more experience doing just this. Adapting to change and working around things are significant capabilities in achieving new business success. If you are an older person considering starting a new business, think about the things you can do better than most other people - think about your strengths and use them. business start-ups for younger people Younger people can be very successful starting new businesses just as much as older people can be. The essential principle of playing to your strengths applies, although the implications are different for younger people compared to older people. Younger people are likely to have lots of fresh ideas. This is an advantage, so avoid people pour cold water on them. Test your ideas on potential customers, rather than to take advice from those people who are ready with their buckets of water. Next, get the help you need. Its difficult for young people to know all the answers. Youll have the ideas and the energy to make things happen, but consider the gaps in your experience, and the things you dont enjoy doing, and seek good quality reliable help for these things. Getting good help at what you cant do or dont want to do will enable you to put all your energy into what you are good at and what you want to spend your time doing. Young people sometimes try to force themselves to fit into roles or responsibilities that are not comfortable or natural. This is de-stabilising and stressful. Learn what you love and excel at, and focus on building success from this. Which brings us back to playing to your strengths. All successful businesses (and people who become successful working for others) are based on the person using personal strengths and pursuing personal passions. Success in business is always based on doing something you love and enjoy, which is fundamentally related to your natural strengths and unique personal potential, whatever that is. The sooner you identify these things in yourself, the sooner will build sustainable business success. planning business success - in summary Spreadsheets, mission statements, planning templates and other process elements of new business creation and development are tools. They enable the business to be properly structured, started and run. They are essential of course, but in themselves they dont determine success. Business success is determined by deeper factors. Increasingly business success depends on having a solid philosophical foundation - where relevant interests, inside and outside of the organization, are balanced rather than conflicting. The bigger the business, the more widely it must consider how it relates to external interests and responsibilities - to society and the world at large. A business with this sort of harmony and balance built into its shape and principles at the outset has a huge advantage over a business which contains tensions and competing pressures. Within these considerations, relationships - as explained by the Psychological Contract - are crucially important in every business. Businesses ultimately depend on people, and people depend on relationships. Aside from this - and without diminishing the significance of other vital business components such as reliability, value, quality, etc. which are necessary merely to survive at a basic level - uniqueness and passion are the remaining special ingredients for success: Uniqueness (just one word, with so many implications) - so that people will want what you offer, and Passion, so that you will enjoy being and offering your best - and so that this belief and commitment conveys to others. authorshipreferencingCrowdsourcing is a very popular means of obtaining the large amounts of labeled data that modern machine learning methods require. Although cheap and fast to obtain, crowdsourced labels suffer from significant amounts of error, thereby degrading the performance of downstream machine learning tasks. With the goal of improving the quality of the labeled data, we seek to mitigate the many errors that occur due to silly mistakes or inadvertent errors by crowdsourcing workers. We propose a two-stage setting for crowdsourcing where the worker first answers the questions, and is then allowed to change her answers after looking at a (noisy) reference answer. We mathematically formulate this process and develop mechanisms to incentivize workers to act appropriately. Our mathematical guarantees show that our mechanism incentivizes the workers to answer honestly in both stages, and refrain from answering randomly in the first stage or simply copying in the second. Numerical experiments reveal a significant boost in performance that such 8220self-correction8221 can provide when using crowdsourcing to train machine learning algorithms. There are various parametric models for analyzing pairwise comparison data, including the Bradley-Terry-Luce (BTL) and Thurstone models, but their reliance on strong parametric assumptions is limiting. In this work, we study a flexible model for pairwise comparisons, under which the probabilities of outcomes are required only to satisfy a natural form of stochastic transitivity. This class includes parametric models including the BTL and Thurstone models as special cases, but is considerably more general. We provide various examples of models in this broader stochastically transitive class for which classical parametric models provide poor fits. Despite this greater flexibility, we show that the matrix of probabilities can be estimated at the same rate as in standard parametric models. On the other hand, unlike in the BTL and Thurstone models, computing the minimax-optimal estimator in the stochastically transitive model is non-trivial, and we explore various computationally tractable alternatives. We show that a simple singular value thresholding algorithm is statistically consistent but does not achieve the minimax rate. We then propose and study algorithms that achieve the minimax rate over interesting sub-classes of the full stochastically transitive class. We complement our theoretical results with thorough numerical simulations. We show how any binary pairwise model may be uprooted to a fully symmetric model, wherein original singleton potentials are transformed to potentials on edges to an added variable, and then rerooted to a new model on the original number of variables. The new model is essentially equivalent to the original model, with the same partition function and allowing recovery of the original marginals or a MAP conguration, yet may have very different computational properties that allow much more efficient inference. This meta-approach deepens our understanding, may be applied to any existing algorithm to yield improved methods in practice, generalizes earlier theoretical results, and reveals a remarkable interpretation of the triplet-consistent polytope. We show how deep learning methods can be applied in the context of crowdsourcing and unsupervised ensemble learning. First, we prove that the popular model of Dawid and Skene, which assumes that all classifiers are conditionally independent, is to a Restricted Boltzmann Machine (RBM) with a single hidden node. Hence, under this model, the posterior probabilities of the true labels can be instead estimated via a trained RBM. Next, to address the more general case, where classifiers may strongly violate the conditional independence assumption, we propose to apply RBM-based Deep Neural Net (DNN). Experimental results on various simulated and real-world datasets demonstrate that our proposed DNN approach outperforms other state-of-the-art methods, in particular when the data violates the conditional independence assumption. Revisiting Semi-Supervised Learning with Graph Embeddings Zhilin Yang Carnegie Mellon University . William Cohen CMU . Ruslan Salakhudinov U. of Toronto Paper AbstractWe present a semi-supervised learning framework based on graph embeddings. Given a graph between instances, we train an embedding for each instance to jointly predict the class label and the neighborhood context in the graph. We develop both transductive and inductive variants of our method. In the transductive variant of our method, the class labels are determined by both the learned embeddings and input feature vectors, while in the inductive variant, the embeddings are defined as a parametric function of the feature vectors, so predictions can be made on instances not seen during training. On a large and diverse set of benchmark tasks, including text classification, distantly supervised entity extraction, and entity classification, we show improved performance over many of the existing models. Reinforcement learning can acquire complex behaviors from high-level specifications. However, defining a cost function that can be optimized effectively and encodes the correct task is challenging in practice. We explore how inverse optimal control (IOC) can be used to learn behaviors from demonstrations, with applications to torque control of high-dimensional robotic systems. Our method addresses two key challenges in inverse optimal control: first, the need for informative features and effective regularization to impose structure on the cost, and second, the difficulty of learning the cost function under unknown dynamics for high-dimensional continuous systems. To address the former challenge, we present an algorithm capable of learning arbitrary nonlinear cost functions, such as neural networks, without meticulous feature engineering. To address the latter challenge, we formulate an efficient sample-based approximation for MaxEnt IOC. We evaluate our method on a series of simulated tasks and real-world robotic manipulation problems, demonstrating substantial improvement over prior methods both in terms of task complexity and sample efficiency. In learning latent variable models (LVMs), it is important to effectively capture infrequent patterns and shrink model size without sacrificing modeling power. Various studies have been done to 8220diversify8221 a LVM, which aim to learn a diverse set of latent components in LVMs. Most existing studies fall into a frequentist-style regularization framework, where the components are learned via point estimation. In this paper, we investigate how to 8220diversify8221 LVMs in the paradigm of Bayesian learning, which has advantages complementary to point estimation, such as alleviating overfitting via model averaging and quantifying uncertainty. We propose two approaches that have complementary advantages. One is to define diversity-promoting mutual angular priors which assign larger density to components with larger mutual angles based on Bayesian network and von Mises-Fisher distribution and use these priors to affect the posterior via Bayes rule. We develop two efficient approximate posterior inference algorithms based on variational inference and Markov chain Monte Carlo sampling. The other approach is to impose diversity-promoting regularization directly over the post-data distribution of components. These two methods are applied to the Bayesian mixture of experts model to encourage the 8220experts8221 to be diverse and experimental results demonstrate the effectiveness and efficiency of our methods. High dimensional nonparametric regression is an inherently difficult problem with known lower bounds depending exponentially in dimension. A popular strategy to alleviate this curse of dimensionality has been to use additive models of emph , which model the regression function as a sum of independent functions on each dimension. Though useful in controlling the variance of the estimate, such models are often too restrictive in practical settings. Between non-additive models which often have large variance and first order additive models which have large bias, there has been little work to exploit the trade-off in the middle via additive models of intermediate order. In this work, we propose salsa, which bridges this gap by allowing interactions between variables, but controls model capacity by limiting the order of interactions. salsas minimises the residual sum of squares with squared RKHS norm penalties. Algorithmically, it can be viewed as Kernel Ridge Regression with an additive kernel. When the regression function is additive, the excess risk is only polynomial in dimension. Using the Girard-Newton formulae, we efficiently sum over a combinatorial number of terms in the additive expansion. Via a comparison on 15 real datasets, we show that our method is competitive against 21 other alternatives. We propose an extension to Hawkes processes by treating the levels of self-excitation as a stochastic differential equation. Our new point process allows better approximation in application domains where events and intensities accelerate each other with correlated levels of contagion. We generalize a recent algorithm for simulating draws from Hawkes processes whose levels of excitation are stochastic processes, and propose a hybrid Markov chain Monte Carlo approach for model fitting. Our sampling procedure scales linearly with the number of required events and does not require stationarity of the point process. A modular inference procedure consisting of a combination between Gibbs and Metropolis Hastings steps is put forward. We recover expectation maximization as a special case. Our general approach is illustrated for contagion following geometric Brownian motion and exponential Langevin dynamics. Rank aggregation systems collect ordinal preferences from individuals to produce a global ranking that represents the social preference. To reduce the computational complexity of learning the global ranking, a common practice is to use rank-breaking. Individuals preferences are broken into pairwise comparisons and then applied to efficient algorithms tailored for independent pairwise comparisons. However, due to the ignored dependencies, naive rank-breaking approaches can result in inconsistent estimates. The key idea to produce unbiased and accurate estimates is to treat the paired comparisons outcomes unequally, depending on the topology of the collected data. In this paper, we provide the optimal rank-breaking estimator, which not only achieves consistency but also achieves the best error bound. This allows us to characterize the fundamental tradeoff between accuracy and complexity in some canonical scenarios. Further, we identify how the accuracy depends on the spectral gap of a corresponding comparison graph. Dropout distillation Samuel Rota Bul FBK . Lorenzo Porzi FBK . Peter Kontschieder Microsoft Research Cambridge Paper AbstractDropout is a popular stochastic regularization technique for deep neural networks that works by randomly dropping (i. e. zeroing) units from the network during training. This randomization process allows to implicitly train an ensemble of exponentially many networks sharing the same parametrization, which should be averaged at test time to deliver the final prediction. A typical workaround for this intractable averaging operation consists in scaling the layers undergoing dropout randomization. This simple rule called 8216standard dropout8217 is efficient, but might degrade the accuracy of the prediction. In this work we introduce a novel approach, coined 8216dropout distillation8217, that allows us to train a predictor in a way to better approximate the intractable, but preferable, averaging process, while keeping under control its computational efficiency. We are thus able to construct models that are as efficient as standard dropout, or even more efficient, while being more accurate. Experiments on standard benchmark datasets demonstrate the validity of our method, yielding consistent improvements over conventional dropout. Metadata-conscious anonymous messaging Giulia Fanti UIUC . Peter Kairouz UIUC . Sewoong Oh UIUC . Kannan Ramchandran UC Berkeley . Pramod Viswanath UIUC Paper AbstractAnonymous messaging platforms like Whisper and Yik Yak allow users to spread messages over a network (e. g. a social network) without revealing message authorship to other users. The spread of messages on these platforms can be modeled by a diffusion process over a graph. Recent advances in network analysis have revealed that such diffusion processes are vulnerable to author deanonymization by adversaries with access to metadata, such as timing information. In this work, we ask the fundamental question of how to propagate anonymous messages over a graph to make it difficult for adversaries to infer the source. In particular, we study the performance of a message propagation protocol called adaptive diffusion introduced in (Fanti et al. 2015). We prove that when the adversary has access to metadata at a fraction of corrupted graph nodes, adaptive diffusion achieves asymptotically optimal source-hiding and significantly outperforms standard diffusion. We further demonstrate empirically that adaptive diffusion hides the source effectively on real social networks. The Teaching Dimension of Linear Learners Ji Liu University of Rochester . Xiaojin Zhu University of Wisconsin . Hrag Ohannessian University of Wisconsin-Madison Paper AbstractTeaching dimension is a learning theoretic quantity that specifies the minimum training set size to teach a target model to a learner. Previous studies on teaching dimension focused on version-space learners which maintain all hypotheses consistent with the training data, and cannot be applied to modern machine learners which select a specific hypothesis via optimization. This paper presents the first known teaching dimension for ridge regression, support vector machines, and logistic regression. We also exhibit optimal training sets that match these teaching dimensions. Our approach generalizes to other linear learners. Truthful Univariate Estimators Ioannis Caragiannis University of Patras . Ariel Procaccia Carnegie Mellon University . Nisarg Shah Carnegie Mellon University Paper AbstractWe revisit the classic problem of estimating the population mean of an unknown single-dimensional distribution from samples, taking a game-theoretic viewpoint. In our setting, samples are supplied by strategic agents, who wish to pull the estimate as close as possible to their own value. In this setting, the sample mean gives rise to manipulation opportunities, whereas the sample median does not. Our key question is whether the sample median is the best (in terms of mean squared error) truthful estimator of the population mean. We show that when the underlying distribution is symmetric, there are truthful estimators that dominate the median. Our main result is a characterization of worst-case optimal truthful estimators, which provably outperform the median, for possibly asymmetric distributions with bounded support. Why Regularized Auto-Encoders learn Sparse Representation Devansh Arpit SUNY Buffalo . Yingbo Zhou SUNY Buffalo . Hung Ngo SUNY Buffalo . Venu Govindaraju SUNY Buffalo Paper AbstractSparse distributed representation is the key to learning useful features in deep learning algorithms, because not only it is an efficient mode of data representation, but also 8212 more importantly 8212 it captures the generation process of most real world data. While a number of regularized auto-encoders (AE) enforce sparsity explicitly in their learned representation and others don8217t, there has been little formal analysis on what encourages sparsity in these models in general. Our objective is to formally study this general problem for regularized auto-encoders. We provide sufficient conditions on both regularization and activation functions that encourage sparsity. We show that multiple popular models (de-noising and contractive auto encoders, e. g.) and activations (rectified linear and sigmoid, e. g.) satisfy these conditions thus, our conditions help explain sparsity in their learned representation. Thus our theoretical and empirical analysis together shed light on the properties of regularizationactivation that are conductive to sparsity and unify a number of existing auto-encoder models and activation functions under the same analytical framework. k-variates: more pluses in the k-means Richard Nock Nicta 038 ANU . Raphael Canyasse Ecole Polytechnique and The Technion . Roksana Boreli Data61 . Frank Nielsen Ecole Polytechnique and Sony CS Labs Inc. Paper Abstractk-means seeding has become a de facto standard for hard clustering algorithms. In this paper, our first contribution is a two-way generalisation of this seeding, k-variates, that includes the sampling of general densities rather than just a discrete set of Dirac densities anchored at the point locations, textit a generalisation of the well known Arthur-Vassilvitskii (AV) approximation guarantee, in the form of a textit approximation bound of the textit optimum. This approximation exhibits a reduced dependency on the 8220noise8221 component with respect to the optimal potential 8212 actually approaching the statistical lower bound. We show that k-variates textit to efficient (biased seeding) clustering algorithms tailored to specific frameworks these include distributed, streaming and on-line clustering, with textit approximation results for these algorithms. Finally, we present a novel application of k-variates to differential privacy. For either the specific frameworks considered here, or for the differential privacy setting, there is little to no prior results on the direct application of k-means and its approximation bounds 8212 state of the art contenders appear to be significantly more complex and or display less favorable (approximation) properties. We stress that our algorithms can still be run in cases where there is textit closed form solution for the population minimizer. We demonstrate the applicability of our analysis via experimental evaluation on several domains and settings, displaying competitive performances vs state of the art. Multi-Player Bandits 8212 a Musical Chairs Approach Jonathan Rosenski Weizmann Institute of Science . Ohad Shamir Weizmann Institute of Science . Liran Szlak Weizmann Institute of Science Paper AbstractWe consider a variant of the stochastic multi-armed bandit problem, where multiple players simultaneously choose from the same set of arms and may collide, receiving no reward. This setting has been motivated by problems arising in cognitive radio networks, and is especially challenging under the realistic assumption that communication between players is limited. We provide a communication-free algorithm (Musical Chairs) which attains constant regret with high probability, as well as a sublinear-regret, communication-free algorithm (Dynamic Musical Chairs) for the more difficult setting of players dynamically entering and leaving throughout the game. Moreover, both algorithms do not require prior knowledge of the number of players. To the best of our knowledge, these are the first communication-free algorithms with these types of formal guarantees. The Information Sieve Greg Ver Steeg Information Sciences Institute . Aram Galstyan Information Sciences Institute Paper AbstractWe introduce a new framework for unsupervised learning of representations based on a novel hierarchical decomposition of information. Intuitively, data is passed through a series of progressively fine-grained sieves. Each layer of the sieve recovers a single latent factor that is maximally informative about multivariate dependence in the data. The data is transformed after each pass so that the remaining unexplained information trickles down to the next layer. Ultimately, we are left with a set of latent factors explaining all the dependence in the original data and remainder information consisting of independent noise. We present a practical implementation of this framework for discrete variables and apply it to a variety of fundamental tasks in unsupervised learning including independent component analysis, lossy and lossless compression, and predicting missing values in data. Deep Speech 2. End-to-End Speech Recognition in English and Mandarin Dario Amodei . Rishita Anubhai . Eric Battenberg . Carl Case . Jared Casper . Bryan Catanzaro . JingDong Chen . Mike Chrzanowski Baidu USA, Inc. . Adam Coates . Greg Diamos Baidu USA, Inc. . Erich Elsen Baidu USA, Inc. . Jesse Engel . Linxi Fan . Christopher Fougner . Awni Hannun Baidu USA, Inc. . Billy Jun . Tony Han . Patrick LeGresley . Xiangang Li Baidu . Libby Lin . Sharan Narang . Andrew Ng . Sherjil Ozair . Ryan Prenger . Sheng Qian Baidu . Jonathan Raiman . Sanjeev Satheesh Baidu SVAIL . David Seetapun . Shubho Sengupta . Chong Wang . Yi Wang . Zhiqian Wang . Bo Xiao . Yan Xie Baidu . Dani Yogatama . Jun Zhan . zhenyao Zhu Paper AbstractWe show that an end-to-end deep learning approach can be used to recognize either English or Mandarin Chinese speechtwo vastly different languages. Because it replaces entire pipelines of hand-engineered components with neural networks, end-to-end learning allows us to handle a diverse variety of speech including noisy environments, accents and different languages. Key to our approach is our application of HPC techniques, enabling experiments that previously took weeks to now run in days. This allows us to iterate more quickly to identify superior architectures and algorithms. As a result, in several cases, our system is competitive with the transcription of human workers when benchmarked on standard datasets. Finally, using a technique called Batch Dispatch with GPUs in the data center, we show that our system can be inexpensively deployed in an online setting, delivering low latency when serving users at scale. An important question in feature selection is whether a selection strategy recovers the 8220true8221 set of features, given enough data. We study this question in the context of the popular Least Absolute Shrinkage and Selection Operator (Lasso) feature selection strategy. In particular, we consider the scenario when the model is misspecified so that the learned model is linear while the underlying real target is nonlinear. Surprisingly, we prove that under certain conditions, Lasso is still able to recover the correct features in this case. We also carry out numerical studies to empirically verify the theoretical results and explore the necessity of the conditions under which the proof holds. We propose minimum regret search (MRS), a novel acquisition function for Bayesian optimization. MRS bears similarities with information-theoretic approaches such as entropy search (ES). However, while ES aims in each query at maximizing the information gain with respect to the global maximum, MRS aims at minimizing the expected simple regret of its ultimate recommendation for the optimum. While empirically ES and MRS perform similar in most of the cases, MRS produces fewer outliers with high simple regret than ES. We provide empirical results both for a synthetic single-task optimization problem as well as for a simulated multi-task robotic control problem. CryptoNets: Applying Neural Networks to Encrypted Data with High Throughput and Accuracy Ran Gilad-Bachrach Microsoft Research . Nathan Dowlin Princeton . Kim Laine Microsoft Research . Kristin Lauter Microsoft Research . Michael Naehrig Microsoft Research . John Wernsing Microsoft Research Paper AbstractApplying machine learning to a problem which involves medical, financial, or other types of sensitive data, not only requires accurate predictions but also careful attention to maintaining data privacy and security. Legal and ethical requirements may prevent the use of cloud-based machine learning solutions for such tasks. In this work, we will present a method to convert learned neural networks to CryptoNets, neural networks that can be applied to encrypted data. This allows a data owner to send their data in an encrypted form to a cloud service that hosts the network. The encryption ensures that the data remains confidential since the cloud does not have access to the keys needed to decrypt it. Nevertheless, we will show that the cloud service is capable of applying the neural network to the encrypted data to make encrypted predictions, and also return them in encrypted form. These encrypted predictions can be sent back to the owner of the secret key who can decrypt them. Therefore, the cloud service does not gain any information about the raw data nor about the prediction it made. We demonstrate CryptoNets on the MNIST optical character recognition tasks. CryptoNets achieve 99 accuracy and can make around 59000 predictions per hour on a single PC. Therefore, they allow high throughput, accurate, and private predictions. Spectral methods for dimensionality reduction and clustering require solving an eigenproblem defined by a sparse affinity matrix. When this matrix is large, one seeks an approximate solution. The standard way to do this is the Nystrom method, which first solves a small eigenproblem considering only a subset of landmark points, and then applies an out-of-sample formula to extrapolate the solution to the entire dataset. We show that by constraining the original problem to satisfy the Nystrom formula, we obtain an approximation that is computationally simple and efficient, but achieves a lower approximation error using fewer landmarks and less runtime. We also study the role of normalization in the computational cost and quality of the resulting solution. As a widely used non-linear activation, Rectified Linear Unit (ReLU) separates noise and signal in a feature map by learning a threshold or bias. However, we argue that the classification of noise and signal not only depends on the magnitude of responses, but also the context of how the feature responses would be used to detect more abstract patterns in higher layers. In order to output multiple response maps with magnitude in different ranges for a particular visual pattern, existing networks employing ReLU and its variants have to learn a large number of redundant filters. In this paper, we propose a multi-bias non-linear activation (MBA) layer to explore the information hidden in the magnitudes of responses. It is placed after the convolution layer to decouple the responses to a convolution kernel into multiple maps by multi-thresholding magnitudes, thus generating more patterns in the feature space at a low computational cost. It provides great flexibility of selecting responses to different visual patterns in different magnitude ranges to form rich representations in higher layers. Such a simple and yet effective scheme achieves the state-of-the-art performance on several benchmarks. We propose a novel multi-task learning method that can minimize the effect of negative transfer by allowing asymmetric transfer between the tasks based on task relatedness as well as the amount of individual task losses, which we refer to as Asymmetric Multi-task Learning (AMTL). To tackle this problem, we couple multiple tasks via a sparse, directed regularization graph, that enforces each task parameter to be reconstructed as a sparse combination of other tasks, which are selected based on the task-wise loss. We present two different algorithms to solve this joint learning of the task predictors and the regularization graph. The first algorithm solves for the original learning objective using alternative optimization, and the second algorithm solves an approximation of it using curriculum learning strategy, that learns one task at a time. We perform experiments on multiple datasets for classification and regression, on which we obtain significant improvements in performance over the single task learning and symmetric multitask learning baselines. This paper illustrates a novel approach to the estimation of generalization error of decision tree classifiers. We set out the study of decision tree errors in the context of consistency analysis theory, which proved that the Bayes error can be achieved only if when the number of data samples thrown into each leaf node goes to infinity. For the more challenging and practical case where the sample size is finite or small, a novel sampling error term is introduced in this paper to cope with the small sample problem effectively and efficiently. Extensive experimental results show that the proposed error estimate is superior to the well known K-fold cross validation methods in terms of robustness and accuracy. Moreover it is orders of magnitudes more efficient than cross validation methods. We study the convergence properties of the VR-PCA algorithm introduced by cite for fast computation of leading singular vectors. We prove several new results, including a formal analysis of a block version of the algorithm, and convergence from random initialization. We also make a few observations of independent interest, such as how pre-initializing with just a single exact power iteration can significantly improve the analysis, and what are the convexity and non-convexity properties of the underlying optimization problem. We consider the problem of principal component analysis (PCA) in a streaming stochastic setting, where our goal is to find a direction of approximate maximal variance, based on a stream of i. i.d. data points in realsd. A simple and computationally cheap algorithm for this is stochastic gradient descent (SGD), which incrementally updates its estimate based on each new data point. However, due to the non-convex nature of the problem, analyzing its performance has been a challenge. In particular, existing guarantees rely on a non-trivial eigengap assumption on the covariance matrix, which is intuitively unnecessary. In this paper, we provide (to the best of our knowledge) the first eigengap-free convergence guarantees for SGD in the context of PCA. This also partially resolves an open problem posed in cite . Moreover, under an eigengap assumption, we show that the same techniques lead to new SGD convergence guarantees with better dependence on the eigengap. Dealbreaker: A Nonlinear Latent Variable Model for Educational Data Andrew Lan Rice University . Tom Goldstein University of Maryland . Richard Baraniuk Rice University . Christoph Studer Cornell University Paper AbstractStatistical models of student responses on assessment questions, such as those in homeworks and exams, enable educators and computer-based personalized learning systems to gain insights into students knowledge using machine learning. Popular student-response models, including the Rasch model and item response theory models, represent the probability of a student answering a question correctly using an affine function of latent factors. While such models can accurately predict student responses, their ability to interpret the underlying knowledge structure (which is certainly nonlinear) is limited. In response, we develop a new, nonlinear latent variable model that we call the dealbreaker model, in which a students success probability is determined by their weakest concept mastery. We develop efficient parameter inference algorithms for this model using novel methods for nonconvex optimization. We show that the dealbreaker model achieves comparable or better prediction performance as compared to affine models with real-world educational datasets. We further demonstrate that the parameters learned by the dealbreaker model are interpretablethey provide key insights into which concepts are critical (i. e. the dealbreaker) to answering a question correctly. We conclude by reporting preliminary results for a movie-rating dataset, which illustrate the broader applicability of the dealbreaker model. We derive a new discrepancy statistic for measuring differences between two probability distributions based on combining Stein8217s identity and the reproducing kernel Hilbert space theory. We apply our result to test how well a probabilistic model fits a set of observations, and derive a new class of powerful goodness-of-fit tests that are widely applicable for complex and high dimensional distributions, even for those with computationally intractable normalization constants. Both theoretical and empirical properties of our methods are studied thoroughly. Variable Elimination in the Fourier Domain Yexiang Xue Cornell University . Stefano Ermon . Ronan Le Bras Cornell University . Carla . Bart Paper AbstractThe ability to represent complex high dimensional probability distributions in a compact form is one of the key insights in the field of graphical models. Factored representations are ubiquitous in machine learning and lead to major computational advantages. We explore a different type of compact representation based on discrete Fourier representations, complementing the classical approach based on conditional independencies. We show that a large class of probabilistic graphical models have a compact Fourier representation. This theoretical result opens up an entirely new way of approximating a probability distribution. We demonstrate the significance of this approach by applying it to the variable elimination algorithm. Compared with the traditional bucket representation and other approximate inference algorithms, we obtain significant improvements. Low-rank matrix approximation has been widely adopted in machine learning applications with sparse data, such as recommender systems. However, the sparsity of the data, incomplete and noisy, introduces challenges to the algorithm stability 8212 small changes in the training data may significantly change the models. As a result, existing low-rank matrix approximation solutions yield low generalization performance, exhibiting high error variance on the training dataset, and minimizing the training error may not guarantee error reduction on the testing dataset. In this paper, we investigate the algorithm stability problem of low-rank matrix approximations. We present a new algorithm design framework, which (1) introduces new optimization objectives to guide stable matrix approximation algorithm design, and (2) solves the optimization problem to obtain stable low-rank approximation solutions with good generalization performance. Experimental results on real-world datasets demonstrate that the proposed work can achieve better prediction accuracy compared with both state-of-the-art low-rank matrix approximation methods and ensemble methods in recommendation task. Given samples from two densities p and q, density ratio estimation (DRE) is the problem of estimating the ratio pq. Two popular discriminative approaches to DRE are KL importance estimation (KLIEP), and least squares importance fitting (LSIF). In this paper, we show that KLIEP and LSIF both employ class-probability estimation (CPE) losses. Motivated by this, we formally relate DRE and CPE, and demonstrate the viability of using existing losses from one problem for the other. For the DRE problem, we show that essentially any CPE loss (eg logistic, exponential) can be used, as this equivalently minimises a Bregman divergence to the true density ratio. We show how different losses focus on accurately modelling different ranges of the density ratio, and use this to design new CPE losses for DRE. For the CPE problem, we argue that the LSIF loss is useful in the regime where one wishes to rank instances with maximal accuracy at the head of the ranking. In the course of our analysis, we establish a Bregman divergence identity that may be of independent interest. We study nonconvex finite-sum problems and analyze stochastic variance reduced gradient (SVRG) methods for them. SVRG and related methods have recently surged into prominence for convex optimization given their edge over stochastic gradient descent (SGD) but their theoretical analysis almost exclusively assumes convexity. In contrast, we prove non-asymptotic rates of convergence (to stationary points) of SVRG for nonconvex optimization, and show that it is provably faster than SGD and gradient descent. We also analyze a subclass of nonconvex problems on which SVRG attains linear convergence to the global optimum. We extend our analysis to mini-batch variants of SVRG, showing (theoretical) linear speedup due to minibatching in parallel settings. Hierarchical Variational Models Rajesh Ranganath . Dustin Tran Columbia University . Blei David Columbia Paper AbstractBlack box variational inference allows researchers to easily prototype and evaluate an array of models. Recent advances allow such algorithms to scale to high dimensions. However, a central question remains: How to specify an expressive variational distribution that maintains efficient computation To address this, we develop hierarchical variational models (HVMs). HVMs augment a variational approximation with a prior on its parameters, which allows it to capture complex structure for both discrete and continuous latent variables. The algorithm we develop is black box, can be used for any HVM, and has the same computational efficiency as the original approximation. We study HVMs on a variety of deep discrete latent variable models. HVMs generalize other expressive variational distributions and maintains higher fidelity to the posterior. The field of mobile health (mHealth) has the potential to yield new insights into health and behavior through the analysis of continuously recorded data from wearable health and activity sensors. In this paper, we present a hierarchical span-based conditional random field model for the key problem of jointly detecting discrete events in such sensor data streams and segmenting these events into high-level activity sessions. Our model includes higher-order cardinality factors and inter-event duration factors to capture domain-specific structure in the label space. We show that our model supports exact MAP inference in quadratic time via dynamic programming, which we leverage to perform learning in the structured support vector machine framework. We apply the model to the problems of smoking and eating detection using four real data sets. Our results show statistically significant improvements in segmentation performance relative to a hierarchical pairwise CRF. Binary embeddings with structured hashed projections Anna Choromanska Courant Institute, NYU . Krzysztof Choromanski Google Research NYC . Mariusz Bojarski NVIDIA . Tony Jebara Columbia . Sanjiv Kumar . Yann Paper AbstractWe consider the hashing mechanism for constructing binary embeddings, that involves pseudo-random projections followed by nonlinear (sign function) mappings. The pseudorandom projection is described by a matrix, where not all entries are independent random variables but instead a fixed budget of randomness is distributed across the matrix. Such matrices can be efficiently stored in sub-quadratic or even linear space, provide reduction in randomness usage (i. e. number of required random values), and very often lead to computational speed ups. We prove several theoretical results showing that projections via various structured matrices followed by nonlinear mappings accurately preserve the angular distance between input high-dimensional vectors. To the best of our knowledge, these results are the first that give theoretical ground for the use of general structured matrices in the nonlinear setting. In particular, they generalize previous extensions of the Johnson - Lindenstrauss lemma and prove the plausibility of the approach that was so far only heuristically confirmed for some special structured matrices. Consequently, we show that many structured matrices can be used as an efficient information compression mechanism. Our findings build a better understanding of certain deep architectures, which contain randomly weighted and untrained layers, and yet achieve high performance on different learning tasks. We empirically verify our theoretical findings and show the dependence of learning via structured hashed projections on the performance of neural network as well as nearest neighbor classifier. A Variational Analysis of Stochastic Gradient Algorithms Stephan Mandt Columbia University . Matthew Hoffman Adobe Research . Blei David Columbia Paper AbstractStochastic Gradient Descent (SGD) is an important algorithm in machine learning. With constant learning rates, it is a stochastic process that, after an initial phase of convergence, generates samples from a stationary distribution. We show that SGD with constant rates can be effectively used as an approximate posterior inference algorithm for probabilistic modeling. Specifically, we show how to adjust the tuning parameters of SGD such as to match the resulting stationary distribution to the posterior. This analysis rests on interpreting SGD as a continuous-time stochastic process and then minimizing the Kullback-Leibler divergence between its stationary distribution and the target posterior. (This is in the spirit of variational inference.) In more detail, we model SGD as a multivariate Ornstein-Uhlenbeck process and then use properties of this process to derive the optimal parameters. This theoretical framework also connects SGD to modern scalable inference algorithms we analyze the recently proposed stochastic gradient Fisher scoring under this perspective. We demonstrate that SGD with properly chosen constant rates gives a new way to optimize hyperparameters in probabilistic models. This paper proposes a new mechanism for sampling training instances for stochastic gradient descent (SGD) methods by exploiting any side-information associated with the instances (for e. g. class-labels) to improve convergence. Previous methods have either relied on sampling from a distribution defined over training instances or from a static distribution that fixed before training. This results in two problems a) any distribution that is set apriori is independent of how the optimization progresses and b) maintaining a distribution over individual instances could be infeasible in large-scale scenarios. In this paper, we exploit the side information associated with the instances to tackle both problems. More specifically, we maintain a distribution over classes (instead of individual instances) that is adaptively estimated during the course of optimization to give the maximum reduction in the variance of the gradient. Intuitively, we sample more from those regions in space that have a textit gradient contribution. Our experiments on highly multiclass datasets show that our proposal converge significantly faster than existing techniques. Tensor regression has shown to be advantageous in learning tasks with multi-directional relatedness. Given massive multiway data, traditional methods are often too slow to operate on or suffer from memory bottleneck. In this paper, we introduce subsampled tensor projected gradient to solve the problem. Our algorithm is impressively simple and efficient. It is built upon projected gradient method with fast tensor power iterations, leveraging randomized sketching for further acceleration. Theoretical analysis shows that our algorithm converges to the correct solution in fixed number of iterations. The memory requirement grows linearly with the size of the problem. We demonstrate superior empirical performance on both multi-linear multi-task learning and spatio-temporal applications. This paper presents a novel distributed variational inference framework that unifies many parallel sparse Gaussian process regression (SGPR) models for scalable hyperparameter learning with big data. To achieve this, our framework exploits a structure of correlated noise process model that represents the observation noises as a finite realization of a high-order Gaussian Markov random process. By varying the Markov order and covariance function for the noise process model, different variational SGPR models result. This consequently allows the correlation structure of the noise process model to be characterized for which a particular variational SGPR model is optimal. We empirically evaluate the predictive performance and scalability of the distributed variational SGPR models unified by our framework on two real-world datasets. Online Stochastic Linear Optimization under One-bit Feedback Lijun Zhang Nanjing University . Tianbao Yang University of Iowa . Rong Jin Alibaba Group . Yichi Xiao Nanjing University . Zhi-hua Zhou Paper AbstractIn this paper, we study a special bandit setting of online stochastic linear optimization, where only one-bit of information is revealed to the learner at each round. This problem has found many applications including online advertisement and online recommendation. We assume the binary feedback is a random variable generated from the logit model, and aim to minimize the regret defined by the unknown linear function. Although the existing method for generalized linear bandit can be applied to our problem, the high computational cost makes it impractical for real-world applications. To address this challenge, we develop an efficient online learning algorithm by exploiting particular structures of the observation model. Specifically, we adopt online Newton step to estimate the unknown parameter and derive a tight confidence region based on the exponential concavity of the logistic loss. Our analysis shows that the proposed algorithm achieves a regret bound of O(dsqrt ), which matches the optimal result of stochastic linear bandits. We present an adaptive online gradient descent algorithm to solve online convex optimization problems with long-term constraints, which are constraints that need to be satisfied when accumulated over a finite number of rounds T, but can be violated in intermediate rounds. For some user-defined trade-off parameter beta in (0, 1), the proposed algorithm achieves cumulative regret bounds of O(Tmax ) and O(T ), respectively for the loss and the constraint violations. Our results hold for convex losses, can handle arbitrary convex constraints and rely on a single computationally efficient algorithm. Our contributions improve over the best known cumulative regret bounds of Mahdavi et al. (2012), which are respectively O(T12) and O(T34) for general convex domains, and respectively O(T23) and O(T23) when the domain is further restricted to be a polyhedral set. We supplement the analysis with experiments validating the performance of our algorithm in practice. Motivated by an application of eliciting users8217 preferences, we investigate the problem of learning hemimetrics, i. e. pairwise distances among a set of n items that satisfy triangle inequalities and non-negativity constraints. In our application, the (asymmetric) distances quantify private costs a user incurs when substituting one item by another. We aim to learn these distances (costs) by asking the users whether they are willing to switch from one item to another for a given incentive offer. Without exploiting structural constraints of the hemimetric polytope, learning the distances between each pair of items requires Theta(n2) queries. We propose an active learning algorithm that substantially reduces this sample complexity by exploiting the structural constraints on the version space of hemimetrics. Our proposed algorithm achieves provably-optimal sample complexity for various instances of the task. For example, when the items are embedded into K tight clusters, the sample complexity of our algorithm reduces to O(n K). Extensive experiments on a restaurant recommendation data set support the conclusions of our theoretical analysis. We present an approach for learning simple algorithms such as copying, multi-digit addition and single digit multiplication directly from examples. Our framework consists of a set of interfaces, accessed by a controller. Typical interfaces are 1-D tapes or 2-D grids that hold the input and output data. For the controller, we explore a range of neural network-based models which vary in their ability to abstract the underlying algorithm from training instances and generalize to test examples with many thousands of digits. The controller is trained using Q-learning with several enhancements and we show that the bottleneck is in the capabilities of the controller rather than in the search incurred by Q-learning. Learning Physical Intuition of Block Towers by Example Adam Lerer Facebook AI Research . Sam Gross Facebook AI Research . Rob Fergus Facebook AI Research Paper AbstractWooden blocks are a common toy for infants, allowing them to develop motor skills and gain intuition about the physical behavior of the world. In this paper, we explore the ability of deep feed-forward models to learn such intuitive physics. Using a 3D game engine, we create small towers of wooden blocks whose stability is randomized and render them collapsing (or remaining upright). This data allows us to train large convolutional network models which can accurately predict the outcome, as well as estimating the trajectories of the blocks. The models are also able to generalize in two important ways: (i) to new physical scenarios, e. g. towers with an additional block and (ii) to images of real wooden blocks, where it obtains a performance comparable to human subjects. Structure Learning of Partitioned Markov Networks Song Liu The Inst. of Stats. Mathe. . Taiji Suzuki . Masashi Sugiyama University of Tokyo . Kenji Fukumizu The Institute of Statistical Mathematics Paper AbstractWe learn the structure of a Markov Network between two groups of random variables from joint observations. Since modelling and learning the full MN structure may be hard, learning the links between two groups directly may be a preferable option. We introduce a novel concept called the emph whose factorization directly associates with the Markovian properties of random variables across two groups. A simple one-shot convex optimization procedure is proposed for learning the emph factorizations of the partitioned ratio and it is theoretically guaranteed to recover the correct inter-group structure under mild conditions. The performance of the proposed method is experimentally compared with the state of the art MN structure learning methods using ROC curves. Real applications on analyzing bipartisanship in US congress and pairwise DNAtime-series alignments are also reported. This work focuses on dynamic regret of online convex optimization that compares the performance of online learning to a clairvoyant who knows the sequence of loss functions in advance and hence selects the minimizer of the loss function at each step. By assuming that the clairvoyant moves slowly (i. e. the minimizers change slowly), we present several improved variation-based upper bounds of the dynamic regret under the true and noisy gradient feedback, which are in light of the presented lower bounds. The key to our analysis is to explore a regularity metric that measures the temporal changes in the clairvoyant8217s minimizers, to which we refer as path variation. Firstly, we present a general lower bound in terms of the path variation, and then show that under full information or gradient feedback we are able to achieve an optimal dynamic regret. Secondly, we present a lower bound with noisy gradient feedback and then show that we can achieve optimal dynamic regrets under a stochastic gradient feedback and two-point bandit feedback. Moreover, for a sequence of smooth loss functions that admit a small variation in the gradients, our dynamic regret under the two-point bandit feedback matches that is achieved with full information. Beyond CCA: Moment Matching for Multi-View Models Anastasia Podosinnikova INRIA 8211 ENS . Francis Bach Inria . Simon Lacoste-Julien INRIA Paper AbstractWe introduce three novel semi-parametric extensions of probabilistic canonical correlation analysis with identifiability guarantees. We consider moment matching techniques for estimation in these models. For that, by drawing explicit links between the new models and a discrete version of independent component analysis (DICA), we first extend the DICA cumulant tensors to the new discrete version of CCA. By further using a close connection with independent component analysis, we introduce generalized covariance matrices, which can replace the cumulant tensors in the moment matching framework, and, therefore, improve sample complexity and simplify derivations and algorithms significantly. As the tensor power method or orthogonal joint diagonalization are not applicable in the new setting, we use non-orthogonal joint diagonalization techniques for matching the cumulants. We demonstrate performance of the proposed models and estimation techniques on experiments with both synthetic and real datasets. We present two computationally inexpensive techniques for estimating the numerical rank of a matrix, combining powerful tools from computational linear algebra. These techniques exploit three key ingredients. The first is to approximate the projector on the non-null invariant subspace of the matrix by using a polynomial filter. Two types of filters are discussed, one based on Hermite interpolation and the other based on Chebyshev expansions. The second ingredient employs stochastic trace estimators to compute the rank of this wanted eigen-projector, which yields the desired rank of the matrix. In order to obtain a good filter, it is necessary to detect a gap between the eigenvalues that correspond to noise and the relevant eigenvalues that correspond to the non-null invariant subspace. The third ingredient of the proposed approaches exploits the idea of spectral density, popular in physics, and the Lanczos spectroscopic method to locate this gap. Unsupervised Deep Embedding for Clustering Analysis Junyuan Xie University of Washington . Ross Girshick Facebook . Ali Farhadi University of Washington Paper AbstractClustering is central to many data-driven application domains and has been studied extensively in terms of distance functions and grouping algorithms. Relatively little work has focused on learning representations for clustering. In this paper, we propose Deep Embedded Clustering (DEC), a method that simultaneously learns feature representations and cluster assignments using deep neural networks. DEC learns a mapping from the data space to a lower-dimensional feature space in which it iteratively optimizes a clustering objective. Our experimental evaluations on image and text corpora show significant improvement over state-of-the-art methods. Dimensionality reduction is a popular approach for dealing with high dimensional data that leads to substantial computational savings. Random projections are a simple and effective method for universal dimensionality reduction with rigorous theoretical guarantees. In this paper, we theoretically study the problem of differentially private empirical risk minimization in the projected subspace (compressed domain). Empirical risk minimization (ERM) is a fundamental technique in statistical machine learning that forms the basis for various learning algorithms. Starting from the results of Chaudhuri et al. (NIPS 2009, JMLR 2011), there is a long line of work in designing differentially private algorithms for empirical risk minimization problems that operate in the original data space. We ask: is it possible to design differentially private algorithms with small excess risk given access to only projected data In this paper, we answer this question in affirmative, by showing that for the class of generalized linear functions, we can obtain excess risk bounds of O(w(Theta) n ) under eps-differential privacy, and O((w(Theta)n) ) under (eps, delta)-differential privacy, given only the projected data and the projection matrix. Here n is the sample size and w(Theta) is the Gaussian width of the parameter space that we optimize over. Our strategy is based on adding noise for privacy in the projected subspace and then lifting the solution to original space by using high-dimensional estimation techniques. A simple consequence of these results is that, for a large class of ERM problems, in the traditional setting (i. e. with access to the original data), under eps-differential privacy, we improve the worst-case risk bounds of Bassily et al. (FOCS 2014). We consider the maximum likelihood parameter estimation problem for a generalized Thurstone choice model, where choices are from comparison sets of two or more items. We provide tight characterizations of the mean square error, as well as necessary and sufficient conditions for correct classification when each item belongs to one of two classes. These results provide insights into how the estimation accuracy depends on the choice of a generalized Thurstone choice model and the structure of comparison sets. We find that for a priori unbiased structures of comparisons, e. g. when comparison sets are drawn independently and uniformly at random, the number of observations needed to achieve a prescribed estimation accuracy depends on the choice of a generalized Thurstone choice model. For a broad set of generalized Thurstone choice models, which includes all popular instances used in practice, the estimation error is shown to be largely insensitive to the cardinality of comparison sets. On the other hand, we found that there exist generalized Thurstone choice models for which the estimation error decreases much faster with the cardinality of comparison sets. Large-Margin Softmax Loss for Convolutional Neural Networks Weiyang Liu Peking University . Yandong Wen South China University of Technology . Zhiding Yu Carnegie Mellon University . Meng Yang Shenzhen University Paper AbstractCross-entropy loss together with softmax is arguably one of the most common used supervision components in convolutional neural networks (CNNs). Despite its simplicity, popularity and excellent performance, the component does not explicitly encourage discriminative learning of features. In this paper, we propose a generalized large-margin softmax (L-Softmax) loss which explicitly encourages intra-class compactness and inter-class separability between learned features. Moreover, L-Softmax not only can adjust the desired margin but also can avoid overfitting. We also show that the L-Softmax loss can be optimized by typical stochastic gradient descent. Extensive experiments on four benchmark datasets demonstrate that the deeply-learned features with L-softmax loss become more discriminative, hence significantly boosting the performance on a variety of visual classification and verification tasks. A Random Matrix Approach to Echo-State Neural Networks Romain Couillet CentraleSupelec . Gilles Wainrib ENS Ulm, Paris, France . Hafiz Tiomoko Ali CentraleSupelec, Gif-sur-Yvette, France . Harry Sevi ENS Lyon, Lyon, Paris Paper AbstractRecurrent neural networks, especially in their linear version, have provided many qualitative insights on their performance under different configurations. This article provides, through a novel random matrix framework, the quantitative counterpart of these performance results, specifically in the case of echo-state networks. Beyond mere insights, our approach conveys a deeper understanding on the core mechanism under play for both training and testing. One-hot CNN (convolutional neural network) has been shown to be effective for text categorization (Johnson 038 Zhang, 2015). We view it as a special case of a general framework which jointly trains a linear model with a non-linear feature generator consisting of text region embedding pooling8217. Under this framework, we explore a more sophisticated region embedding method using Long Short-Term Memory (LSTM). LSTM can embed text regions of variable (and possibly large) sizes, whereas the region size needs to be fixed in a CNN. We seek effective and efficient use of LSTM for this purpose in the supervised and semi-supervised settings. The best results were obtained by combining region embeddings in the form of LSTM and convolution layers trained on unlabeled data. The results indicate that on this task, embeddings of text regions, which can convey complex concepts, are more useful than embeddings of single words in isolation. We report performances exceeding the previous best results on four benchmark datasets. Crowdsourcing systems are popular for solving large-scale labelling tasks with low-paid (or even non-paid) workers. We study the problem of recovering the true labels from noisy crowdsourced labels under the popular Dawid-Skene model. To address this inference problem, several algorithms have recently been proposed, but the best known guarantee is still significantly larger than the fundamental limit. We close this gap under a simple but canonical scenario where each worker is assigned at most two tasks. In particular, we introduce a tighter lower bound on the fundamental limit and prove that Belief Propagation (BP) exactly matches this lower bound. The guaranteed optimality of BP is the strongest in the sense that it is information-theoretically impossible for any other algorithm to correctly la - bel a larger fraction of the tasks. In the general setting, when more than two tasks are assigned to each worker, we establish the dominance result on BP that it outperforms other existing algorithms with known provable guarantees. Experimental results suggest that BP is close to optimal for all regimes considered, while existing state-of-the-art algorithms exhibit suboptimal performances. Learning control has become an appealing alternative to the derivation of control laws based on classic control theory. However, a major shortcoming of learning control is the lack of performance guarantees which prevents its application in many real-world scenarios. As a step in this direction, we provide a stability analysis tool for controllers acting on dynamics represented by Gaussian processes (GPs). We consider arbitrary Markovian control policies and system dynamics given as (i) the mean of a GP, and (ii) the full GP distribution. For the first case, our tool finds a state space region, where the closed-loop system is provably stable. In the second case, it is well known that infinite horizon stability guarantees cannot exist. Instead, our tool analyzes finite time stability. Empirical evaluations on simulated benchmark problems support our theoretical results. Learning a classifier from private data distributed across multiple parties is an important problem that has many potential applications. How can we build an accurate and differentially private global classifier by combining locally-trained classifiers from different parties, without access to any partys private data We propose to transfer the knowledge of the local classifier ensemble by first creating labeled data from auxiliary unlabeled data, and then train a global differentially private classifier. We show that majority voting is too sensitive and therefore propose a new risk weighted by class probabilities estimated from the ensemble. Relative to a non-private solution, our private solution has a generalization error bounded by O(epsilon M ). This allows strong privacy without performance loss when the number of participating parties M is large, such as in crowdsensing applications. We demonstrate the performance of our framework with realistic tasks of activity recognition, network intrusion detection, and malicious URL detection. Network Morphism Tao Wei University at Buffalo . Changhu Wang Microsoft Research . Yong Rui Microsoft Research . Chang Wen Chen Paper AbstractWe present a systematic study on how to morph a well-trained neural network to a new one so that its network function can be completely preserved. We define this as network morphism in this research. After morphing a parent network, the child network is expected to inherit the knowledge from its parent network and also has the potential to continue growing into a more powerful one with much shortened training time. The first requirement for this network morphism is its ability to handle diverse morphing types of networks, including changes of depth, width, kernel size, and even subnet. To meet this requirement, we first introduce the network morphism equations, and then develop novel morphing algorithms for all these morphing types for both classic and convolutional neural networks. The second requirement is its ability to deal with non-linearity in a network. We propose a family of parametric-activation functions to facilitate the morphing of any continuous non-linear activation neurons. Experimental results on benchmark datasets and typical neural networks demonstrate the effectiveness of the proposed network morphism scheme. Second-order optimization methods such as natural gradient descent have the potential to speed up training of neural networks by correcting for the curvature of the loss function. Unfortunately, the exact natural gradient is impractical to compute for large models, and most approximations either require an expensive iterative procedure or make crude approximations to the curvature. We present Kronecker Factors for Convolution (KFC), a tractable approximation to the Fisher matrix for convolutional networks based on a structured probabilistic model for the distribution over backpropagated derivatives. Similarly to the recently proposed Kronecker-Factored Approximate Curvature (K-FAC), each block of the approximate Fisher matrix decomposes as the Kronecker product of small matrices, allowing for efficient inversion. KFC captures important curvature information while still yielding comparably efficient updates to stochastic gradient descent (SGD). We show that the updates are invariant to commonly used reparameterizations, such as centering of the activations. In our experiments, approximate natural gradient descent with KFC was able to train convolutional networks several times faster than carefully tuned SGD. Furthermore, it was able to train the networks in 10-20 times fewer iterations than SGD, suggesting its potential applicability in a distributed setting. Budget constrained optimal design of experiments is a classical problem in statistics. Although the optimal design literature is very mature, few efficient strategies are available when these design problems appear in the context of sparse linear models commonly encountered in high dimensional machine learning and statistics. In this work, we study experimental design for the setting where the underlying regression model is characterized by a ell1-regularized linear function. We propose two novel strategies: the first is motivated geometrically whereas the second is algebraic in nature. We obtain tractable algorithms for this problem and also hold for a more general class of sparse linear models. We perform an extensive set of experiments, on benchmarks and a large multi-site neuroscience study, showing that the proposed models are effective in practice. The latter experiment suggests that these ideas may play a small role in informing enrollment strategies for similar scientific studies in the short-to-medium term future. Minding the Gaps for Block Frank-Wolfe Optimization of Structured SVMs Anton Osokin . Jean-Baptiste Alayrac ENS . Isabella Lukasewitz INRIA . Puneet Dokania INRIA and Ecole Centrale Paris . Simon Lacoste-Julien INRIA Paper AbstractIn this paper, we propose several improvements on the block-coordinate Frank-Wolfe (BCFW) algorithm from Lacoste-Julien et al. (2013) recently used to optimize the structured support vector machine (SSVM) objective in the context of structured prediction, though it has wider applications. The key intuition behind our improvements is that the estimates of block gaps maintained by BCFW reveal the block suboptimality that can be used as an adaptive criterion. First, we sample objects at each iteration of BCFW in an adaptive non-uniform way via gap-based sampling. Second, we incorporate pairwise and away-step variants of Frank-Wolfe into the block-coordinate setting. Third, we cache oracle calls with a cache-hit criterion based on the block gaps. Fourth, we provide the first method to compute an approximate regularization path for SSVM. Finally, we provide an exhaustive empirical evaluation of all our methods on four structured prediction datasets. Exact Exponent in Optimal Rates for Crowdsourcing Chao Gao Yale University . Yu Lu Yale University . Dengyong Zhou Microsoft Research Paper AbstractCrowdsourcing has become a popular tool for labeling large datasets. This paper studies the optimal error rate for aggregating crowdsourced labels provided by a collection of amateur workers. Under the Dawid-Skene probabilistic model, we establish matching upper and lower bounds with an exact exponent mI(pi), where m is the number of workers and I(pi) is the average Chernoff information that characterizes the workers8217 collective ability. Such an exact characterization of the error exponent allows us to state a precise sample size requirement m ge frac logfrac in order to achieve an epsilon misclassification error. In addition, our results imply optimality of various forms of EM algorithms given accurate initializers of the model parameters. Unsupervised learning and supervised learning are key research topics in deep learning. However, as high-capacity supervised neural networks trained with a large amount of labels have achieved remarkable success in many computer vision tasks, the availability of large-scale labeled images reduced the significance of unsupervised learning. Inspired by the recent trend toward revisiting the importance of unsupervised learning, we investigate joint supervised and unsupervised learning in a large-scale setting by augmenting existing neural networks with decoding pathways for reconstruction. First, we demonstrate that the intermediate activations of pretrained large-scale classification networks preserve almost all the information of input images except a portion of local spatial details. Then, by end-to-end training of the entire augmented architecture with the reconstructive objective, we show improvement of the network performance for supervised tasks. We evaluate several variants of autoencoders, including the recently proposed 8220what-where8221 autoencoder that uses the encoder pooling switches, to study the importance of the architecture design. Taking the 16-layer VGGNet trained under the ImageNet ILSVRC 2012 protocol as a strong baseline for image classification, our methods improve the validation-set accuracy by a noticeable margin. (LRR) has been a significant method for segmenting data that are generated from a union of subspaces. It is also known that solving LRR is challenging in terms of time complexity and memory footprint, in that the size of the nuclear norm regularized matrix is n-by-n (where n is the number of samples). In this paper, we thereby develop a novel online implementation of LRR that reduces the memory cost from O(n2) to O(pd), with p being the ambient dimension and d being some estimated rank (d 20 reduction in the model size without any loss in accuracy on CIFAR-10 benchmark. We also demonstrate that fine-tuning can further enhance the accuracy of fixed point DCNs beyond that of the original floating point model. In doing so, we report a new state-of-the-art fixed point performance of 6.78 error-rate on CIFAR-10 benchmark. Provable Algorithms for Inference in Topic Models Sanjeev Arora Princeton University . Rong Ge . Frederic Koehler Princeton University . Tengyu Ma Princeton University . Ankur Moitra Paper AbstractRecently, there has been considerable progress on designing algorithms with provable guarantees 8212typically using linear algebraic methods8212for parameter learning in latent variable models. Designing provable algorithms for inference has proved more difficult. Here we take a first step towards provable inference in topic models. We leverage a property of topic models that enables us to construct simple linear estimators for the unknown topic proportions that have small variance, and consequently can work with short documents. Our estimators also correspond to finding an estimate around which the posterior is well-concentrated. We show lower bounds that for shorter documents it can be information theoretically impossible to find the hidden topics. Finally, we give empirical results that demonstrate that our algorithm works on realistic topic models. It yields good solutions on synthetic data and runs in time comparable to a single iteration of Gibbs sampling. This paper develops an approach for efficiently solving general convex optimization problems specified as disciplined convex programs (DCP), a common general-purpose modeling framework. Specifically we develop an algorithm based upon fast epigraph projections, projections onto the epigraph of a convex function, an approach closely linked to proximal operator methods. We show that by using these operators, we can solve any disciplined convex program without transforming the problem to a standard cone form, as is done by current DCP libraries. We then develop a large library of efficient epigraph projection operators, mirroring and extending work on fast proximal algorithms, for many common convex functions. Finally, we evaluate the performance of the algorithm, and show it often achieves order of magnitude speedups over existing general-purpose optimization solvers. We study the fixed design segmented regression problem: Given noisy samples from a piecewise linear function f, we want to recover f up to a desired accuracy in mean-squared error. Previous rigorous approaches for this problem rely on dynamic programming (DP) and, while sample efficient, have running time quadratic in the sample size. As our main contribution, we provide new sample near-linear time algorithms for the problem that 8211 while not being minimax optimal 8211 achieve a significantly better sample-time tradeoff on large datasets compared to the DP approach. Our experimental evaluation shows that, compared with the DP approach, our algorithms provide a convergence rate that is only off by a factor of 2 to 4, while achieving speedups of three orders of magnitude. Energetic Natural Gradient Descent Philip Thomas CMU . Bruno Castro da Silva . Christoph Dann Carnegie Mellon University . Emma Paper AbstractWe propose a new class of algorithms for minimizing or maximizing functions of parametric probabilistic models. These new algorithms are natural gradient algorithms that leverage more information than prior methods by using a new metric tensor in place of the commonly used Fisher information matrix. This new metric tensor is derived by computing directions of steepest ascent where the distance between distributions is measured using an approximation of energy distance (as opposed to Kullback-Leibler divergence, which produces the Fisher information matrix), and so we refer to our new ascent direction as the energetic natural gradient. Partition Functions from Rao-Blackwellized Tempered Sampling David Carlson Columbia University . Patrick Stinson Columbia University . Ari Pakman Columbia University . Liam Paper AbstractPartition functions of probability distributions are important quantities for model evaluation and comparisons. We present a new method to compute partition functions of complex and multimodal distributions. Such distributions are often sampled using simulated tempering, which augments the target space with an auxiliary inverse temperature variable. Our method exploits the multinomial probability law of the inverse temperatures, and provides estimates of the partition function in terms of a simple quotient of Rao-Blackwellized marginal inverse temperature probability estimates, which are updated while sampling. We show that the method has interesting connections with several alternative popular methods, and offers some significant advantages. In particular, we empirically find that the new method provides more accurate estimates than Annealed Importance Sampling when calculating partition functions of large Restricted Boltzmann Machines (RBM) moreover, the method is sufficiently accurate to track training and validation log-likelihoods during learning of RBMs, at minimal computational cost. In this paper we address the identifiability and efficient learning problems of finite mixtures of Plackett-Luce models for rank data. We prove that for any kgeq 2, the mixture of k Plackett-Luce models for no more than 2k-1 alternatives is non-identifiable and this bound is tight for k2. For generic identifiability, we prove that the mixture of k Plackett-Luce models over m alternatives is if kleqlfloorfrac 2rfloor. We also propose an efficient generalized method of moments (GMM) algorithm to learn the mixture of two Plackett-Luce models and show that the algorithm is consistent. Our experiments show that our GMM algorithm is significantly faster than the EMM algorithm by Gormley 038 Murphy (2008), while achieving competitive statistical efficiency. The combinatorial explosion that plagues planning and reinforcement learning (RL) algorithms can be moderated using state abstraction. Prohibitively large task representations can be condensed such that essential information is preserved, and consequently, solutions are tractably computable. However, exact abstractions, which treat only fully-identical situations as equivalent, fail to present opportunities for abstraction in environments where no two situations are exactly alike. In this work, we investigate approximate state abstractions, which treat nearly-identical situations as equivalent. We present theoretical guarantees of the quality of behaviors derived from four types of approximate abstractions. Additionally, we empirically demonstrate that approximate abstractions lead to reduction in task complexity and bounded loss of optimality of behavior in a variety of environments. Power of Ordered Hypothesis Testing Lihua Lei Lihua . William Fithian UC Berkeley, Department of Statistics Paper AbstractOrdered testing procedures are multiple testing procedures that exploit a pre-specified ordering of the null hypotheses, from most to least promising. We analyze and compare the power of several recent proposals using the asymptotic framework of Li 038 Barber (2015). While accumulation tests including ForwardStop can be quite powerful when the ordering is very informative, they are asymptotically powerless when the ordering is weaker. By contrast, Selective SeqStep, proposed by Barber 038 Candes (2015), is much less sensitive to the quality of the ordering. We compare the power of these procedures in different regimes, concluding that Selective SeqStep dominates accumulation tests if either the ordering is weak or non-null hypotheses are sparse or weak. Motivated by our asymptotic analysis, we derive an improved version of Selective SeqStep which we call Adaptive SeqStep, analogous to Storeys improvement on the Benjamini-Hochberg proce - dure. We compare these methods using the GEO-Query data set analyzed by (Li 038 Barber, 2015) and find Adaptive SeqStep has favorable performance for both good and bad prior orderings. PHOG: Probabilistic Model for Code Pavol Bielik ETH Zurich . Veselin Raychev ETH Zurich . Martin Vechev ETH Zurich Paper AbstractWe introduce a new generative model for code called probabilistic higher order grammar (PHOG). PHOG generalizes probabilistic context free grammars (PCFGs) by allowing conditioning of a production rule beyond the parent non-terminal, thus capturing rich contexts relevant to programs. Even though PHOG is more powerful than a PCFG, it can be learned from data just as efficiently. We trained a PHOG model on a large JavaScript code corpus and show that it is more precise than existing models, while similarly fast. As a result, PHOG can immediately benefit existing programming tools based on probabilistic models of code. We consider the problem of online prediction in changing environments. In this framework the performance of a predictor is evaluated as the loss relative to an arbitrarily changing predictor, whose individual components come from a base class of predictors. Typical results in the literature consider different base classes (experts, linear predictors on the simplex, etc.) separately. Introducing an arbitrary mapping inside the mirror decent algorithm, we provide a framework that unifies and extends existing results. As an example, we prove new shifting regret bounds for matrix prediction problems. Hyperparameter selection generally relies on running multiple full training trials, with selection based on validation set performance. We propose a gradient-based approach for locally adjusting hyperparameters during training of the model. Hyperparameters are adjusted so as to make the model parameter gradients, and hence updates, more advantageous for the validation cost. We explore the approach for tuning regularization hyperparameters and find that in experiments on MNIST, SVHN and CIFAR-10, the resulting regularization levels are within the optimal regions. The additional computational cost depends on how frequently the hyperparameters are trained, but the tested scheme adds only 30 computational overhead regardless of the model size. Since the method is significantly less computationally demanding compared to similar gradient-based approaches to hyperparameter optimization, and consistently finds good hyperparameter values, it can be a useful tool for training neural network models. Many of the recent Trajectory Optimization algorithms alternate between local approximation of the dynamics and conservative policy update. However, linearly approximating the dynamics in order to derive the new policy can bias the update and prevent convergence to the optimal policy. In this article, we propose a new model-free algorithm that backpropagates a local quadratic time-dependent Q-Function, allowing the derivation of the policy update in closed form. Our policy update ensures exact KL-constraint satisfaction without simplifying assumptions on the system dynamics demonstrating improved performance in comparison to related Trajectory Optimization algorithms linearizing the dynamics. Due to its numerous applications, rank aggregation has become a problem of major interest across many fields of the computer science literature. In the vast majority of situations, Kemeny consensus(es) are considered as the ideal solutions. It is however well known that their computation is NP-hard. Many contributions have thus established various results to apprehend this complexity. In this paper we introduce a practical method to predict, for a ranking and a dataset, how close the Kemeny consensus(es) are to this ranking. A major strength of this method is its generality: it does not require any assumption on the dataset nor the ranking. Furthermore, it relies on a new geometric interpretation of Kemeny aggregation that, we believe, could lead to many other results. Horizontally Scalable Submodular Maximization Mario Lucic ETH Zurich . Olivier Bachem ETH Zurich . Morteza Zadimoghaddam Google Research . Andreas Krause Paper AbstractA variety of large-scale machine learning problems can be cast as instances of constrained submodular maximization. Existing approaches for distributed submodular maximization have a critical drawback: The capacity 8211 number of instances that can fit in memory 8211 must grow with the data set size. In practice, while one can provision many machines, the capacity of each machine is limited by physical constraints. We propose a truly scalable approach for distributed submodular maximization under fixed capacity. The proposed framework applies to a broad class of algorithms and constraints and provides theoretical guarantees on the approximation factor for any available capacity. We empirically evaluate the proposed algorithm on a variety of data sets and demonstrate that it achieves performance competitive with the centralized greedy solution. Group Equivariant Convolutional Networks Taco Cohen University of Amsterdam . Max Welling University of Amsterdam CIFAR Paper AbstractWe introduce Group equivariant Convolutional Neural Networks (G-CNNs), a natural generalization of convolutional neural networks that reduces sample complexity by exploiting symmetries. G-CNNs use G-convolutions, a new type of layer that enjoys a substantially higher degree of weight sharing than regular convolution layers. G-convolutions increase the expressive capacity of the network without increasing the number of parameters. Group convolution layers are easy to use and can be implemented with negligible computational overhead for discrete groups generated by translations, reflections and rotations. G-CNNs achieve state of the art results on CIFAR10 and rotated MNIST. The partition function is fundamental for probabilistic graphical models8212it is required for inference, parameter estimation, and model selection. Evaluating this function corresponds to discrete integration, namely a weighted sum over an exponentially large set. This task quickly becomes intractable as the dimensionality of the problem increases. We propose an approximation scheme that, for any discrete graphical model whose parameter vector has bounded norm, estimates the partition function with arbitrarily small error. Our algorithm relies on a near minimax optimal polynomial approximation to the potential function and a Clenshaw-Curtis style quadrature. Furthermore, we show that this algorithm can be randomized to split the computation into a high-complexity part and a low-complexity part, where the latter may be carried out on small computational devices. Experiments confirm that the new randomized algorithm is highly accurate if the parameter norm is small, and is otherwise comparable to methods with unbounded error. Correcting Forecasts with Multifactor Neural Attention Matthew Riemer IBM . Aditya Vempaty IBM . Flavio Calmon IBM . Fenno Heath IBM . Richard Hull IBM . Elham Khabiri IBM Paper AbstractAutomatic forecasting of time series data is a challenging problem in many industries. Current forecast models adopted by businesses do not provide adequate means for including data representing external factors that may have a significant impact on the time series, such as weather, national events, local events, social media trends, promotions, etc. This paper introduces a novel neural network attention mechanism that naturally incorporates data from multiple external sources without the feature engineering needed to get other techniques to work. We demonstrate empirically that the proposed model achieves superior performance for predicting the demand of 20 commodities across 107 stores of one of America8217s largest retailers when compared to other baseline models, including neural networks, linear models, certain kernel methods, Bayesian regression, and decision trees. Our method ultimately accounts for a 23.9 relative improvement as a result of the incorporation of external data sources, and provides an unprecedented level of descriptive ability for a neural network forecasting model. Observational studies are rising in importance due to the widespread accumulation of data in fields such as healthcare, education, employment and ecology. We consider the task of answering counterfactual questions such as, 8220Would this patient have lower blood sugar had she received a different medication8221. We propose a new algorithmic framework for counterfactual inference which brings together ideas from domain adaptation and representation learning. In addition to a theoretical justification, we perform an empirical comparison with previous approaches to causal inference from observational data. Our deep learning algorithm significantly outperforms the previous state-of-the-art. Gaussian Processes (GPs) provide a general and analytically tractable way of modeling complex time-varying, nonparametric functions. The Automatic Bayesian Covariance Discovery (ABCD) system constructs natural-language description of time-series data by treating unknown time-series data nonparametrically using GP with a composite covariance kernel function. Unfortunately, learning a composite covariance kernel with a single time-series data set often results in less informative kernel that may not give qualitative, distinctive descriptions of data. We address this challenge by proposing two relational kernel learning methods which can model multiple time-series data sets by finding common, shared causes of changes. We show that the relational kernel learning methods find more accurate models for regression problems on several real-world data sets US stock data, US house price index data and currency exchange rate data. We introduce a new approach for amortizing inference in directed graphical models by learning heuristic approximations to stochastic inverses, designed specifically for use as proposal distributions in sequential Monte Carlo methods. We describe a procedure for constructing and learning a structured neural network which represents an inverse factorization of the graphical model, resulting in a conditional density estimator that takes as input particular values of the observed random variables, and returns an approximation to the distribution of the latent variables. This recognition model can be learned offline, independent from any particular dataset, prior to performing inference. The output of these networks can be used as automatically-learned high-quality proposal distributions to accelerate sequential Monte Carlo across a diverse range of problem settings. Slice Sampling on Hamiltonian Trajectories Benjamin Bloem-Reddy Columbia University . John Cunningham Columbia University Paper AbstractHamiltonian Monte Carlo and slice sampling are amongst the most widely used and studied classes of Markov Chain Monte Carlo samplers. We connect these two methods and present Hamiltonian slice sampling, which allows slice sampling to be carried out along Hamiltonian trajectories, or transformations thereof. Hamiltonian slice sampling clarifies a class of model priors that induce closed-form slice samplers. More pragmatically, inheriting properties of slice samplers, it offers advantages over Hamiltonian Monte Carlo, in that it has fewer tunable hyperparameters and does not require gradient information. We demonstrate the utility of Hamiltonian slice sampling out of the box on problems ranging from Gaussian process regression to Pitman-Yor based mixture models. Noisy Activation Functions Caglar Glehre . Marcin Moczulski . Misha Denil . Yoshua Bengio U. of Montreal Paper AbstractCommon nonlinear activation functions used in neural networks can cause training difficulties due to the saturation behavior of the activation function, which may hide dependencies that are not visible to vanilla-SGD (using first order gradients only). Gating mechanisms that use softly saturating activation functions to emulate the discrete switching of digital logic circuits are good examples of this. We propose to exploit the injection of appropriate noise so that the gradients may flow easily, even if the noiseless application of the activation function would yield zero gradients. Large noise will dominate the noise-free gradient and allow stochastic gradient descent to explore more. By adding noise only to the problematic parts of the activation function, we allow the optimization procedure to explore the boundary between the degenerate saturating) and the well-behaved parts of the activation function. We also establish connections to simulated annealing, when the amount of noise is annealed down, making it easier to optimize hard objective functions. We find experimentally that replacing such saturating activation functions by noisy variants helps optimization in many contexts, yielding state-of-the-art or competitive results on different datasets and task, especially when training seems to be the most difficult, e. g. when curriculum learning is necessary to obtain good results. PD-Sparse. A Primal and Dual Sparse Approach to Extreme Multiclass and Multilabel Classification Ian En-Hsu Yen University of Texas at Austin . Xiangru Huang UTaustin . Pradeep Ravikumar UT Austin . Kai Zhong ICES department, University of Texas at Austin . Inderjit Paper AbstractWe consider Multiclass and Multilabel classification with extremely large number of classes, of which only few are labeled to each instance. In such setting, standard methods that have training, prediction cost linear to the number of classes become intractable. State-of-the-art methods thus aim to reduce the complexity by exploiting correlation between labels under assumption that the similarity between labels can be captured by structures such as low-rank matrix or balanced tree. However, as the diversity of labels increases in the feature space, structural assumption can be easily violated, which leads to degrade in the testing performance. In this work, we show that a margin-maximizing loss with l1 penalty, in case of Extreme Classification, yields extremely sparse solution both in primal and in dual without sacrificing the expressive power of predictor. We thus propose a Fully-Corrective Block-Coordinate Frank-Wolfe (FC-BCFW) algorithm that exploits both primal and dual sparsity to achieve a complexity sublinear to the number of primal and dual variables. A bi-stochastic search method is proposed to further improve the efficiency. In our experiments on both Multiclass and Multilabel problems, the proposed method achieves significant higher accuracy than existing approaches of Extreme Classification with very competitive training and prediction time.