Algorithms for Irrelevance-Based Partial MAPs

TL;DR

This paper introduces algorithms for computing irrelevance-based partial MAPs in belief networks, providing theoretical properties and modifications to existing algorithms to improve their effectiveness in explanations.

Contribution

It defines two types of irrelevance-based partial MAPs, proves key properties, and adapts the MAP best-first algorithm for efficient computation of these partial MAPs.

Findings

Proved important properties of irrelevance-based partial MAPs.

Modified the standard MAP algorithm to handle these partial MAPs.

Enhanced algorithm effectiveness for domain-independent explanations.

Abstract

Irrelevance-based partial MAPs are useful constructs for domain-independent explanation using belief networks. We look at two definitions for such partial MAPs, and prove important properties that are useful in designing algorithms for computing them effectively. We make use of these properties in modifying our standard MAP best-first algorithm, so as to handle irrelevance-based partial MAPs.