8th Multidisciplinary Workshop on

Advances in Preference Handling

Invited Speaker: Craig Boutilier, University of Toronto (joint work with Tyler Lu)

Title: Scaling Optimization Methods for Data-driven Marketing

The emergence of large-scale, data-driven analytics has greatly improved the ability to predict the behavior of, and the effect of marketing actions on, individual consumers. Indeed, the potential for fully personalized "marketing conversations" is very real. Unfortunately, advances in predictive analytics have significantly outpaced the ability of current decision support tools and optimization algorithms, precisely the tools needed to transform these insights into marketing plans, policies and strategies. This is especially true in large marketing organizations, where large numbers of campaigns, business objectives, product groups, etc. place competing demands on marketing resources - the most important of which may be customer attention.

 In this talk, I will describe a new approach, called dynamic segmentation, for solving large-scale marketing optimization problems. We formulate the problem as a generalized assignment problem (or other mathematical program) and create aggregate segmentations based on both (statistical) predictive models and campaign-specific and organizational objectives. The resulting compression allows problems involving hundreds of campaigns and millions of customers to be solved optimally in tens of milliseconds. I'll briefly describe how the data-intensive components of the algorithm can be distributed to take advantage of modern cluster-computing frameworks, and how the platform supports real-time scenario analysis and re-optimization.


The talks are 15min plus 3min time for questions and answers. The workshop program can also be downloaded as PDF. 

08:30 - 08:40 Welcome

Session 1: Conditional and Constraint-based Preferences

08:40 - 09:00 Counting, Ranking, and Randomly Generating CP-nets. Thomas E. Allen, Judy Goldsmith and Nicholas Mattei

09:00 - 09:20 Constraint-based Preferences via Utility Hyper-graphs. Rafik Hadfi and Takayuki Ito

Session 2: Preference Reasoning

09:20 - 09:40 Manipulation and Bribery in Preference Reasoning under Pareto Principle. Ying Zhu and Miroslaw Truszczynski

09:40 - 10:00 The Likelihood of Structure in Preference Profiles. Marie-Louise Bruner and Martin Lackner

10:00 - 10:20 Aggregating Opinions to Design Energy-Efficient Buildings. Boian Kolev, Samori Price, Sreerag Palangat Veetil, Leandro Soriano Marcolino, Albert Xin Jiang, David Gerber and Milind Tambe

10:30 - 11:00 Coffee Break

Session 3: Voting

11:00 - 11:20 Justified Representation in Approval-Based Committee Voting. Haris Aziz and Toby Walsh

11:20 - 11:40 Controlling elections by replacing candidates: theoretical and experimental results. Andrea Loreggia, Nina Narodytska, Francesca Rossi, Kristen Brent Venable and Toby Walsh

11:40 - 12:00 Computational Aspects of Multi-Winner Approval Voting. Haris Aziz, Serge Gaspers, Joachim Gudmundsson, Simon Mackenzie, Nicholas Mattei and Toby Walsh

12:00 - 12:20 Fixing a Balanced Knockout Tournament.Haris Aziz, Serge Gaspers, Simon Mackenzie, Nicholas Mattei, Paul Stursberg and Toby Walsh 

12:30 - 13:30 Lunch

Invited Talk

13:30 - 14:10 Craig Boutilier: Scaling Optimization Methods for Data-driven Marketing

Session 4: Preference Modeling and Languages

14:10 - 14:30 Modeling Agent's Preferences Based on Prospect Theory. Paulo A. L. Castro and Simon Parsons

14:30 - 14:50 Grouping Queries with SV-Semantics in Preference SQL. Markus Endres, Patrick Roocks, Manuel Huber and Werner Kießling

14:50 - 15:10 A Model for Intransitive Preferences. Sam Saarinen, Craig A. Tovey and Judy Goldsmith

15:10 - 15:30 A Language for Representing and Reasoning about Qualitative Preferences. Xudong Liu and Mirek Truszczynski

15:30 - 16:00 Coffee Break

Session 5

Who is Watching You Eat? J. Goldsmith, N. Mattei and R. Sloan

Uncovering Response Biases in Recommendation. K.-W. Park, B.-H. Kim, T.-S. Park and B.-T. Zhang

Human and Computer Preferences at Chess. K. Regan, T. Biswas and J. Zhou

This web site has been reconstructed from a Wayback Machine recording and may differ in style.