Preferences Improve Learning to Solve Constraint Problems

Smiljana Petrovic, Susan L. Epstein

Search heuristics that embody domain knowledge often reflect preferences among choices. This paper proposes a variety of ways to identify a good mixture of search heuristics and to integrate the preferences they express. Two preference expression methods inspired by the voting literature in political science prove particularly reliable. On hard problems, good methods to integrate such heuristics are shown to speed search, reducing significantly both computation time and the size of the search tree.

Subjects: 15.2 Constraint Satisfaction; 15.7 Search

Submitted: May 14, 2007

This page is copyrighted by AAAI. All rights reserved. Your use of this site constitutes acceptance of all of AAAI's terms and conditions and privacy policy.