A Comparison of Lauritzen-Spiegelhalter, Hugin, and Shenoy-Shafer Architectures for Computing Marginals of Probability Distributions

TL;DR

This paper compares three architectures for exact marginal computation in probabilistic graphical models, focusing on their structure, message-passing schemes, and efficiency to guide optimal architecture choice.

Contribution

It provides a detailed comparison of Lauritzen-Spiegelhalter, Hugin, and Shenoy-Shafer architectures across multiple criteria, highlighting their differences and advantages.

Findings

Hugin architecture offers improved computational efficiency over Lauritzen-Spiegelhalter.

Shenoy-Shafer architecture provides flexible message-passing schemes.

Storage efficiency varies significantly among the three architectures.

Abstract

In the last decade, several architectures have been proposed for exact computation of marginals using local computation. In this paper, we compare three architectures - Lauritzen-Spiegelhalter, Hugin, and Shenoy-Shafer - from the perspective of graphical structure for message propagation, message-passing scheme, computational efficiency, and storage efficiency.