Choice Models and Preference Learning

NIPS workshop, 17 December 2011, Sierra Nevada, Spain

Sponsored by PASCAL 2


Preference learning has been studied for several decades and has drawn increasing attention in recent years due to its importance in web applications, such as ad serving, search, and electronic commerce. In all of these applications, we observe (often discrete) choices that reflect relative preferences among several items, e.g. products, songs, web pages or documents. Moreover, observations are in many cases censored. Hence, the goal is to reconstruct the overall model of preferences by, for example, learning a general ordering function based on the partially observed decisions.

Choice models try to predict the specific choices individuals (or groups of individuals) make when offered a possibly very large number of alternatives. Traditionally, they are concerned with the decision process of individuals and have been studied independently in machine learning, data and web mining, econometrics, and psychology. However, these diverse communities have had few interactions in the past. One goal of this workshop is to foster interdisciplinary exchange, by encouraging abstraction of the underlying problem (and solution) characteristics.

The workshop is motivated by the two following lines of research:

1. Large scale preference learning with sparse data: There has been a great interest and take-up of machine learning techniques for preference learning in learning to rank, information retrieval and recommender systems, as supported by the large proportion of preference learning based literature in the widely regarded conferences such as SIGIR, WSDM, WWW, and CIKM. Different paradigms of machine learning have been further developed and applied to these challenging problems, particularly when there is a large number of users and items but only a small set of user preferences are provided. 

2. Personalization in social networks: recent wide acceptance of social networks has brought great opportunities for services in different domains, thanks to Facebook, LinkedIn, Douban, Twitter, etc. It is important for these service providers to offer personalized service (e.g., personalization of Twitter recommendations).  Social information can improve the inference for user preferences. However, it is still challenging to infer user preferences based on social relationship. 

As such, we especially encourage submissions on theory, methods, and applications focusing on large-scale preference learning and choice models in social media. In order to avoid a dispersed research workshop, we solicit submissions (papers, demos and project descriptions) and participation that specifically tackle the research areas as below:

  • Preference elicitation
  • Ranking aggregation
  • Choice models and inference
  • Statistical relational learning for preferences
  • Link prediction for preferences
  • Learning Structured Preferences
  • Multi-task preference learning 
  • (Social) collaborative filtering


  • 7.30-7:45         Opening                  
  • 7:45-8:30         Invited talk: Online Learning with Implicit User PreferencesThorsten Joachims
  • 8:30-9:00         Contributed talk: Exact Bayesian Pairwise Preference Learning and Inference on the Uniform Convex Polytope, S. Sanner, E. Abbasnejad
  • 9:00-9:15         Coffee break
  • 9:15-10:00       3-minute pitch for posters, Authors with poster papers
  • 10:00-15:30      Coffee break, poster session, lunch, skiing break
  • 15:30-16:15      Invited talk: Collaborative Learning of Preferences for Recommending Games and MediaThore Graepel
  • 16:15-16:45      Contributed talk: Label Ranking with Abstention: Predicting Partial Orders by Thresholding Probability Distributions, Weiwei Cheng and Eyke Huellermeier
  • 16:45-17:00      Coffee break
  • 17:00-17:45      Invited talk: Probabilistic Models for Preference LearningZoubin Ghahramani  
  • 17:45-18:15      Contributed talk:  Approximate Sorting of Preference Data, L. M. Busse, M. H. Chehreghani, J. M. Buhmann
  • 18:15-18:20      Break
  • 18:20-18:50      
    Contributed talk: Selective Sampling with Almost Optimal Guarantees for Learning to Rank from Pairwise Preferences
    N. Ailon, R. Begleiter, E. Ezra.
  • 18:50-19:30      Discussion & Open research problems

Accepted Papers

Multi-Label Classification with Relevance Ordering (pdf )
    by Miao Xu, Yu-Feng Li and Zhi-Hua Zhou

A GP Classification Approach to Preference Learning (pdf )
    by Ferenc Huszar. 

Approximate Sorting of Preference Data (pdf )
    by Ludwig M. Busse, Morteza Haghir Chehreghani and Joachim M. Buhmann

Collaborative Context-aware Preference Learning (pdf )
    by Alexandros Karatzoglou, Linas Baltrunas and Matthias Bohmer

Online Learning with Preference Feedback (pdf )
    by Pannagadatta Shivaswamy and Thorsten Joachims

Sparse Gaussian Processes for Learning Preferences (pdf )
    by Ehsan Abbasnejad, Edwin Bonilla and Scott Sanner. 

On Sparse Multi-Task Gaussian Process Priors for Music Preference Learning (pdf ),
    by Jens Brehm Nielsen, Bjorn Sand Jensen and Jan Larsen. 

Learning Item Trees for Probabilistic Modeling of Implicit Feedback (pdf ),
    by Andriy Mnih and Yee Whye Teh. 

Selective Sampling with Almost Optimal Guarantees for Learning to Rank from Pairwise Preferences (pdf )
    by Nir Ailon, Ron Begleiter and Esther Ezra

Learning to Recommend Links using Graph Structure and Node Content (pdf )
     by Antonino Freno, Gemma C Garriga and Mikaela Keller. 

Active Ranking in Practice: General Ranking Functions with Sample Complexity Bounds (pdf )
    by Kevin Jamieson and Robert Nowak. 

Preference-based Reinforcement Learning (pdf )
    by Riad Akrour, Marc Schoenauer and Michele Sebag.

Label Ranking with Abstention: Predicting Partial Orders by Thresholding Probability Distributions (pdf )
    by Weiwei Cheng and Eyke Huellermeier

Exact Bayesian Pairwise Preference Learning and Inference on the Uniform Convex Polytope (pdf ),
    by Scott Sanner and Ehsan Abbasnejad.

Preference elicitation for interactive learning of Optimization Modulo Theory problems (pdf ),
    by Paolo Campigotto, Andrea Passerini and Roberto Battiti.

Submission Instructions:

We solicit extended abstracts using the NIPS style files (available here), preferably 2 to 4 pages, but no more than 8 pages. Submissions should include the title, authors' names, and email addresses. We will post the final version of the papers on the workshop web page and encourage authors to post their contribution on arXiv

Papers should be submitted to the EasyChair system at

Talks will be published on

Important Dates:

Paper submission deadline: 3 November 2011 (Extended)
Author notification: 5 November 2011
Final paper due: 1 December 2011
Workshop date: 17 December 2011        

Program Committee:
  • Nir Ailon, Israel Institute of Technology
  • Edwin Bonilla, NICTA-ANU
  • Tiberio Caetano, NICTA - ANU
  • François Caron, INRIA
  • Jonathan Chung-Kuan Huang, Carnegie Mellon University
  • Chris Dance, Xerox Research Centre Europe
  • Jo­hannes Fürnkranz, TU Darmstadt
  • John Guiver, Microsoft Research Cambridge
  • Eyke Hüllermeier, Universität Marburg
  • Hang Li, Microsoft Research Asia
  • Robert Nowak, University of Wisconsin-Madison 
  • Filip Radlinski, Microsoft
  • Chu Wei, Yahoo! Labs
  • Markus Weimer, Yahoo! Labs
  • Kai Yu, NEC Labs
  • Zhao Xu, Fraunhofer IAIS

Program Co-Chair:

PASCAL - Pattern Analysis, Statistical Modelling and Computational Learning

Workshop Contact:

Shengbo (dot)
+33 (0)4 76 61 50 47

Xerox Research Centre Europe, 6 chemin de Maupertuis, 38240 Meylan, France

Locations of visitors to this page