Heuristic Search as Evidential Reasoning

TL;DR

This paper introduces BPS, a Bayesian inference-based heuristic search method that outperforms traditional algorithms in problem-solving efficiency, demonstrated through empirical tests on the Eight Puzzle.

Contribution

It presents a novel Bayesian inference mechanism for heuristic search, showing improved performance and computational efficiency over traditional methods.

Findings

BPS outperforms traditional heuristic search algorithms in efficiency.

Empirical tests on the Eight Puzzle demonstrate BPS's superior decision-making.

BPS requires significantly fewer node expansions to achieve better results.

Abstract

BPS, the Bayesian Problem Solver, applies probabilistic inference and decision-theoretic control to flexible, resource-constrained problem-solving. This paper focuses on the Bayesian inference mechanism in BPS, and contrasts it with those of traditional heuristic search techniques. By performing sound inference, BPS can outperform traditional techniques with significantly less computational effort. Empirical tests on the Eight Puzzle show that after only a few hundred node expansions, BPS makes better decisions than does the best existing algorithm after several million node expansions