Constructive Preference Elicitation by Setwise Max-margin Learning

TL;DR

This paper introduces a setwise max-margin learning approach for preference elicitation in large configuration spaces, producing diverse item sets for informative user queries and encouraging sparse utility models.

Contribution

It generalizes max-margin learning to sets, enabling preference elicitation in complex spaces and improving scalability over existing Bayesian methods.

Findings

Outperforms Bayesian preference elicitation methods in scalability and effectiveness.

Produces diverse item sets for more informative user queries.

Encourages sparse utility models focusing on key features.

Abstract

In this paper we propose an approach to preference elicitation that is suitable to large configuration spaces beyond the reach of existing state-of-the-art approaches. Our setwise max-margin method can be viewed as a generalization of max-margin learning to sets, and can produce a set of "diverse" items that can be used to ask informative queries to the user. Moreover, the approach can encourage sparsity in the parameter space, in order to favor the assessment of utility towards combinations of weights that concentrate on just few features. We present a mixed integer linear programming formulation and show how our approach compares favourably with Bayesian preference elicitation alternatives and easily scales to realistic datasets.