Minimax regret based elicitation of generalized additive utilities

TL;DR

This paper introduces a new method for eliciting generalized additive utilities using minimax regret, with novel query strategies that leverage local structure for practical preference-based optimization.

Contribution

It presents a semantic foundation for GAI utility elicitation with minimax regret and proposes new query types and strategies exploiting local structure for computational efficiency.

Findings

Provides a practical approach for preference-based constrained configuration

Enables effective search in multiattribute product databases

Utilizes local GAI structure for computational feasibility

Abstract

We describe the semantic foundations for elicitation of generalized additively independent (GAI) utilities using the minimax regret criterion, and propose several new query types and strategies for this purpose. Computational feasibility is obtained by exploiting the local GAI structure in the model. Our results provide a practical approach for implementing preference-based constrained configuration optimization as well as effective search in multiattribute product databases.