AI summaryⓘ
The authors address the challenge of making predictions and calculations in very complex Bayesian networks, where traditional methods become too slow. They propose breaking the big network into smaller, manageable parts called directed convex subgraphs, organized in a structure named a minimal d-decomposition tree. This lets them work with simpler pieces instead of the whole network at once, which saves computing power and allows multiple calculations to happen at the same time. They also create two new algorithms that use this setup to estimate parameters and make inferences efficiently. Their experiments show this approach is faster than standard methods, especially when focusing on a small number of variables, while still being accurate.
Bayesian networksProbabilistic inferenceJoint distributionDirected convex subgraphsMinimal d-decomposition treeJunction-tree algorithmParameter estimationParallel computationLow-dimensional queries
Abstract
Probabilistic inference in high-dimensional Bayesian networks is difficult because exact manipulation of the joint distribution scales exponentially with network size. We propose a decomposition framework based on directed convex subgraphs and introduce a minimal d-decomposition tree. Together, they provide a principled alternative to classical junction-tree constructions. The proposed framework represents the joint distribution by lower-dimensional sub-models that can be learned and stored separately. This decomposition reduces computational cost and naturally enables parallel computation. Based on a minimal d-decomposition tree, we further develop two parallel algorithms for parameter estimation and probabilistic inference. Experiments show that the proposed method substantially improves computational efficiency over junction-tree methods while maintaining inference accuracy, especially for low-dimensional queries.