Parallelizing Probabilistic Inference: Some Early Explorations

TL;DR

This paper explores the potential for parallelizing belief network inference, finding significant speedup possible mainly through parallelizing conformal product operations on hypercube machines.

Contribution

It provides an experimental analysis of parallelism opportunities in belief network inference, highlighting the importance of factoring and the limited benefit from topological parallelism.

Findings

Substantial speedup achievable via parallel conformal product operations

Parallelism at the topological or clustering level is negligible

Effective parallelization depends on appropriate factoring strategies

Abstract

We report on an experimental investigation into opportunities for parallelism in beliefnet inference. Specifically, we report on a study performed of the available parallelism, on hypercube style machines, of a set of randomly generated belief nets, using factoring (SPI) style inference algorithms. Our results indicate that substantial speedup is available, but that it is available only through parallelization of individual conformal product operations, and depends critically on finding an appropriate factoring. We find negligible opportunity for parallelism at the topological, or clustering tree, level.