Example: stock market

Preference Learning: A Tutorial Introduction

Preference Learning: A Tutorial IntroductionEykeH llermeierKnowledgeEngineering & BioinformaticsLabDept. of Mathematicsand Computer ScienceMarburg University, GermanyPL-10 @ ECML/PKDD 2010, Barcelona, September 24, 2010 Johannes F rnkranzKnowledgeEngineeringDept. ofComputer ScienceTechnical University Darmstadt, GermanyECML/PKDD-2010 Tutorialon PreferenceLearning| Part 1 | J. F rnkranz& E. H llermeierWhatisPreference Learning ? Preference learning is an emerging subfield of machine learning Roughly speaking, it deals with the learning of (predictive) Preference modelsfrom observed (or extracted) Preference information MACHINE LEARNINGPREFERNCE MODELING and DECISION ANALYSISP reference Learningcomputerscienceartificialintelli genceoperationsresearchsocialsciences(vo tingandchoicetheory)economicsanddecision theory2 ECML/PKDD-2010 Tutorialon PreferenceLearning| Part 1 | J.

A Tutorial Introduction Eyke Hüllermeier Knowledge Engineering & Bioinformatics Lab Dept. of Mathematics and Computer Science Marburg University, Germany PL-10 @ ECML/PKDD 2010, Barcelona, September 24, 2010 Johannes Fürnkranz Knowledge Engineering Dept. of Computer Science

Fullscreen Download

Tags:

  Introduction, Tutorials, Tutorial introduction

Information

Domain:

Source:

Link to this page:

Please notify us if you found a problem with this document:

Other abuse

Advertisement

Transcription of Preference Learning: A Tutorial Introduction

1 Preference Learning: A Tutorial IntroductionEykeH llermeierKnowledgeEngineering & BioinformaticsLabDept. of Mathematicsand Computer ScienceMarburg University, GermanyPL-10 @ ECML/PKDD 2010, Barcelona, September 24, 2010 Johannes F rnkranzKnowledgeEngineeringDept. ofComputer ScienceTechnical University Darmstadt, GermanyECML/PKDD-2010 Tutorialon PreferenceLearning| Part 1 | J. F rnkranz& E. H llermeierWhatisPreference Learning ? Preference learning is an emerging subfield of machine learning Roughly speaking, it deals with the learning of (predictive) Preference modelsfrom observed (or extracted) Preference information MACHINE LEARNINGPREFERNCE MODELING and DECISION ANALYSISP reference Learningcomputerscienceartificialintelli genceoperationsresearchsocialsciences(vo tingandchoicetheory)economicsanddecision theory2 ECML/PKDD-2010 Tutorialon PreferenceLearning| Part 1 | J.

2 F rnkranz& E. H llermeierWorkshops and RelatedEvents NIPS 01: New MethodsforPreferenceElicitation NIPS 02: Beyond Classification and Regression: Learning Rankings, Preferences, Equality Predicates, and Other Structures KI 03: Preference Learning: Models, Methods, Applications NIPS 04: Learning With Structured Outputs NIPS 05: Workshop on Learningto Rank IJCAI 05: Advancesin PreferenceHandling SIGIR 07 10: Workshop on Learning to Rank for Information Retrieval ECML/PDKK 08 10: Workshop on Preference Learning NIPS 09: Workshop on Advancesin Ranking American Institute of MathematicsWorkshop in Summer 2010: TheMathematicsof Ranking3 ECML/PKDD-2010 Tutorialon PreferenceLearning| Part 1 | J. F rnkranz& E. H llermeierPreferencesin ArtificialIntelligenceUser preferencesplaya keyrolein variousfieldsof application: recommendersystems, adaptive userinterfaces, adaptive retrievalsystems, autonomousagents(electroniccommerce), games.

3 Preferencesin AI research: Preference representation (CP nets, GAU networks, logical representations, fuzzyconstraints, ..) reasoningwith preferences (decision theory, constraint satisfaction, non-monotonicreasoning, ..) Preference acquisition ( Preference elicitation, Preference learning, ..)Moregenerally, preferences isa keytopicin currentAI research4 ECML/PKDD-2010 Tutorialon PreferenceLearning| Part 1 | J. F rnkranz& E. H Learning Tasks (Eyke) Functions(Johannes) Learning Techniques(Eyke) PreferenceLearning(Johannes) Tutorialon PreferenceLearning| Part 1 | J. F rnkranz& E. H llermeierPreferenceLearningPreference learning problems can be distinguished along several problemdimensions, including representation of preferences, type of Preference model: utility function (ordinal, cardinal), Preference relation (partial order, ranking.)

4 , logical representation, .. description of individuals/users and alternatives/items: identifier, feature vector, structured object, .. type of training input: direct or indirect feedback, complete orincompleterelations, utilities, ..6 ECML/PKDD-2010 Tutorialon PreferenceLearning| Part 1 | J. F rnkranz& E. H llermeierPreference Learning7 Preferencesabsoluterelativebinarygradual total orderpartial order A ++-0 (ordinal) regression classification/rankingECML/PKDD-2010 Tutorialon PreferenceLearning| Part 1 | J. F rnkranz& E. H llermeierStructureof thisOverview(1) Preference Learning as an extensionof conventionalsupervisedlearning: Learna mappingthatmapsinstancesto preferencemodels( structured/complexoutputprediction).(2)O ther settings(objectranking, instanceranking, CF.)

5 8 ECML/PKDD-2010 Tutorialon PreferenceLearning| Part 1 | J. F rnkranz& E. H llermeierStructureof thisOverview(1) Preference Learning as an extensionof conventionalsupervisedlearning: Learna mappingthatmapsinstancesto preferencemodels( structured/complexoutputprediction).Inst ancesaretypically(thoughnotnecessarily) characterizedin termsof a outputspaceconsistsofpreferencemodelsove ra fixedsetofalternatives (classes, labels, ..) representedin termsofan identifier extensionsofmulti-classclassification9 ECML/PKDD-2010 Tutorialon PreferenceLearning| Part 1 | J. F rnkranz& E. H llermeierMultilabel Classification[Tsoumakas& Katakis2007] a fixedsetof items: likedordislikedLOSSECML/PKDD-2010 Tutorialon PreferenceLearning| Part 1 | J. F rnkranz& E. H llermeierMultilabel B DCAB inarypreferenceson a fixedsetof items: likedordislikedA rankingof all itemsLOSSECML/PKDD-2010 Tutorialon PreferenceLearning| Part 1 | J.

6 F rnkranz& E. H llermeierGradedMultilabel Classification[Chenget al. 2010] +++ ++--+ + ++++ +0++ ++++ +0++ ++--+GroundtruthOrdinalpreferenceson a fixedsetof items: likedordislikedA rankingof all itemsLOSSECML/PKDD-2010 Tutorialon PreferenceLearning| Part 1 | J. F rnkranz& E. H llermeierGradedMultilabel +++ ++--+ + ++++ +0++ ++++ ++--+Groundtruth B DCAO rdinalpreferenceson a fixedsetof items: likedordislikedA rankingof all itemsLOSSECML/PKDD-2010 Tutorialon PreferenceLearning| Part 1 | J. F rnkranz& E. H llermeierLabel Ranking [H llermeieret al. 2008] B, B C, C D,A D, C D, A A, C D, A D,A D, A,A B, C B, A B DCAA rankingof all Tutorialon PreferenceLearning| Part 1 | J. F rnkranz& E. H llermeierCalibratedLabel Ranking [F rnkranzet al.]

7 2008]15 Combiningabsolute andrelative evaluation: a bcd efgrelevantpositivelikedirrelevantnegati vedislikedPreferencesabsoluterelativeECM L/PKDD-2010 Tutorialon PreferenceLearning| Part 1 | J. F rnkranz& E. H llermeierStructureof thisOverview(1) Preference Learning as an extensionof conventionalsupervisedlearning: Learna mappingthatmapsinstancesto preferencemodels( structuredoutputprediction).(2)Othersett ingsobjectranking, instanceranking( no outputspace )collaborativefiltering( no inputspace )16 ECML/PKDD-2010 Tutorialon PreferenceLearning| Part 1 | J. F rnkranz& E. H llermeierObjectRanking [Cohen et al. 99]17 TrainingPrediction(rankinga newsetof objects)Groundtruth(rankingortop-ranking orsubsetof relevant objects)Pairwisepreferencesbetweenobject s(instances).

8 ECML/PKDD-2010 Tutorialon PreferenceLearning| Part 1 | J. F rnkranz& E. H llermeierInstanceRanking [F rnkranzet al. 2009] ++.. +Prediction(rankinga newsetof objects)+0++++--+0+--00----Groundtruth(o rdinalclasses)ECML/PKDD-2010 Tutorialon PreferenceLearning| Part 1 | J. F rnkranz& E. H llermeierInstanceRanking [F rnkranzet al. 2009]19predictedranking, , throughsortingbyestimatedscoremostlikely goodmostlikelybadrankingerrorExtension of AUC maximizationto thepolytomouscase, in whichinstancesareratedon an ordinalscalesuch as { bad, medium, good}Query setof instancesto beranked(truelabelsareunknown).ECML/PKDD -2010 Tutorialon PreferenceLearning| Part 1 | J. F rnkranz& E. H llermeierCollaborativeFiltering[Goldberg et al. 1992] : verybad, 2: bad, 3: fair, 4: good, 5: excellentU S E R SP R O D U C T SInputs and outputsas identifiers, absolute preferencesin termsof Tutorialon PreferenceLearning| Part 1 | J.

9 F rnkranz& E. H llermeierPreferenceLearningTasks21taskin putoutputtrainingpredictiongroundtruthco llaborativefilteringidentifieridentifier absoluteordinalabsoluteordinalabsoluteor dinalmultilabelclassificationfeatureiden tifierabsolutebinaryabsolutebinaryabsolu tebinarymultilabelrankingfeatureidentifi erabsolutebinaryrankingabsolutebinarygra dedmultilabelclassificationfeatureidenti fierabsoluteordinalabsoluteordinalabsolu teordinallabelrankingfeatureidentifierre lativebinaryrankingrankingobjectrankingf eature--relativebinaryrankingrankingorsu bsetinstancerankingfeatureidentifierabso luteordinalrankingabsoluteordinalgeneral izedclassificationrankingTwomaindirectio ns: (1) Ranking and variants(2) generalizationsof preferenceinformationECML/PKDD-2010 Tutorialon PreferenceLearning| Part 1 | J.

10 F rnkranz& E. H llermeierBeyondRanking: PredictingPartial Oders [Chevaleyreet al. 2010, Chenget al. 2010b]22 Rankings (stricttotal orders) canbegeneralizedin different ways, , throughindifference(ties) orincomparability Predictingpartial ordersamongalternatives: Learningconditionalpreference(CP) networks Twointerpretations: Partial abstentiondueto uncertainty(targetisa total order) versuspredictionof trulypartial order relation. BarcelonaParisRomeLondon cannotcompare ECML/PKDD-2010 Tutorialon PreferenceLearning| Part 1 | J. F rnkranz& E. H llermeierLossFunctions23absolute utilitydegreeabsolute utilitydegreesubsetof preferreditemssubsetof preferreditemssubsetof preferreditemsrankingof itemsfuzzysubsetof preferreditemsfuzzysubsetof preferreditemsrankingof itemsrankingof itemsrankingof itemsorderedpartitionof itemsThings to becompared:standardcomparisonof scalarpredictionsnon-standardcomparisons ECML/PKDD-2010 Tutorialon PreferenceLearning| Part 1 | J.

Show more

Documents from same domain

Chapter 18, Sections 1{3 - Computer Science

Chapter 18, Sections 1{3 - Computer Science

www2.cs.uh.edu

Chapter 18, Sections 1{3 2. Learning Learning is essential for unknown environments, i.e., when designer lacks omniscience Learning is useful as a system construction method, i.e., expose the agent to reality rather than trying to write it down

  Chapter, Chapter 18

Discrete Mathematics for Computer Science

Discrete Mathematics for Computer Science

www2.cs.uh.edu

2.5.2 Application: DNF and Combinatorial Networks 124 2.5.3 Conjunctive Normal Form 125 2.5.4 Application: CNF and Combinatorial Networks 127 2.5.5 Testing Satisfiability and Validity 127 2.5.6 The Famous 'P Af r Conjecture 129 2.5.7 Resolution Proofs: Automating Logic 129 2.6 Exercises 131 2.7 Predicates and Quantification 134

  Testing, Combinatorial

Test -Driven Development By Example - UH

Test -Driven Development By Example - UH

www2.cs.uh.edu

Test -driven development 10 of 133. Fear . Test -driven development (TDD) is a way of managing fear during programming . I don’t mean fear in a bad way, pow widdle prwogwammew needs a pacifiew, but fear in the legitimate, this -is-a-hard -problem -and -I-can’t -see -the -end -from -the -beginning sense.

  Development, Tests, Driven, Test driven development

Fake Review Detection: Classification and Analysis of Real ...

Fake Review Detection: Classification and Analysis of Real ...

www2.cs.uh.edu

Area Under the ROC ) Curveof 0.78 was reported using logistic regression. The assumption, however, is too restricted for detecting generic fake The work in [reviews. 24] used similar features but applied a co-training method on a manually labeled dataset of fake and non-fake reviews attaining an F1-score of 0.63.

  Analysis

Related documents

IHSDM Tutorial - Introduction

IHSDM Tutorial - Introduction

www.ihsdm.org

IHSDM Tutorial - Introduction Purpose: The purpose of this Tutorial is to provide users with a guided overview of the IHSDM software. Upon completing the self-paced tutorial, the user should have a basic understanding of the IHSDM software and its components, and experience in applying

  Introduction, Tutorials

ENVI Tutorial: Introduction to ENVI

ENVI Tutorial: Introduction to ENVI

faculty.wwu.edu

Tutorial: Introduction to ENVI Overview of This Tutorial This tutorial is designed to introduce you to the basic concepts of the ENVI software and some of its key features.

  Introduction, Tutorials, Vine, Envi tutorial, Introduction to envi

Cortical Neurons and Circuits: A Tutorial Introduction

Cortical Neurons and Circuits: A Tutorial Introduction

www.mrc.uidaho.edu

Cortical Neurons and Circuits: A Tutorial Introduction Richard B. Wells April, 2005 Abstract. This paper is a tutorial review of the structure, composition, and statistical modeling of the organization of the neocortex. It begins with a general overview of the layered structure of

  Introduction, Tutorials, Circuit, A tutorial introduction, And circuits

A Tutorial Introduction to the Lambda Calculus

A Tutorial Introduction to the Lambda Calculus

www.utdallas.edu

A Tutorial Introduction to the Lambda Calculus Raul Rojas FU Berlin, WS-97/98 Abstract This paper is a short and painless introduction to the calculus.

  Introduction, Tutorials, Calculus, Lambda, Tutorial introduction, Lambda calculus

BI Tutorial: Introduction to the Data Warehouse Center

BI Tutorial: Introduction to the Data Warehouse Center

users.sdsc.edu

72 BI Tutorial: Introduction to the Data Warehouse Center All statements regarding IBM’s future direction or intent are subject to change or withdrawal without …

  Introduction, Tutorials

Tutorial On Introduction to 8085 Architecture and …

Tutorial On Introduction to 8085 Architecture and …

ce.sharif.edu

Tutorial On Introduction to 8085 Architecture and Programming . Contents 1. Internal architecture of 8085 microprocessor 2. 8085 system bus 3. 8085 pin description. 4. 8085 functional description. 5. Programming model of 8085 microprocessor 6. Addressing modes. ... The microprocessor uses this register to sequence the execution of the instructions.

  Introduction, Tutorials, 5808, Microprocessor, 8085 microprocessor, Tutorial on introduction to 8085

Introduction to SystemC Tutorial

Introduction to SystemC Tutorial

homepages.cae.wisc.edu

This tutorial is an introduction to the use of C++ for describing hardware models. Essential constructs for the creation of SystemC models are also introduced. Note this is a …

  Introduction, Tutorials, An introduction, Systemc tutorial, Systemc

Lecture 15 Introduction to Survival Analysis

Lecture 15 Introduction to Survival Analysis

www.stat.columbia.edu

Introduction to Survival Analysis BIOST 515 February 26, 2004 BIOST 515, Lecture 15. ... Survival analysis is used to analyze data in which the time until the event is of interest. The response is often referred to as a failure time, survival time, or event time. BIOST 515, Lecture 15 1.

  Analysis, Introduction, Survival, Survival analysis

Introduction to Stan - University of Colorado Boulder

Introduction to Stan - University of Colorado Boulder

bechtel.colorado.edu

Introduction Bayesian Stats About Stan Examples Tips and Tricks Introduction to Stan Cameron Bracken University of Colorado Boulder February 2015

  Introduction, Asnt, Introduction to stan

Introduction to Engineering

Introduction to Engineering

engineering.nyu.edu

Why Engineering in K-12 • Real-world engineering applications and examples concretize complex math and science concepts • Students are engaged in experiential learning • Students’ creativity is challenged, developed, and enhanced • Students’ soft skills in communication and team-work are developed • Students are better equipped for college-level ...

  Introduction, Engineering, Introduction to engineering

Related search queries

Tutorial, Introduction, ENVI Tutorial: Introduction to ENVI, And Circuits: A Tutorial Introduction, Tutorial Introduction, Lambda Calculus, Tutorial On Introduction to 8085, 8085 microprocessor, SystemC Tutorial, An introduction, SystemC, Survival analysis, Introduction to Stan, Introduction to Engineering