The Complexity of Bayesian Network Learning: Revisiting the Superstructure

TL;DR

This paper explores the parameterized complexity of Bayesian Network Structure Learning, revealing new fixed-parameter tractability results based on different graph parameters and input representations, advancing theoretical understanding of the problem.

Contribution

It introduces fixed-parameter tractability results for BNSL based on feedback edge sets and input representations, providing a comprehensive complexity classification.

Findings

BNSL is fixed-parameter tractable when parameterized by feedback edge set size.

Using additive input representation makes BNSL fixed-parameter tractable by treewidth.

The results extend to the related problem of Polytree Learning.

Abstract

We investigate the parameterized complexity of Bayesian Network Structure Learning (BNSL), a classical problem that has received significant attention in empirical but also purely theoretical studies. We follow up on previous works that have analyzed the complexity of BNSL w.r.t. the so-called superstructure of the input. While known results imply that BNSL is unlikely to be fixed-parameter tractable even when parameterized by the size of a vertex cover in the superstructure, here we show that a different kind of parameterization - notably by the size of a feedback edge set - yields fixed-parameter tractability. We proceed by showing that this result can be strengthened to a localized version of the feedback edge set, and provide corresponding lower bounds that complement previous results to provide a complexity classification of BNSL w.r.t. virtually all well-studied graph parameters. We then analyze how the complexity of BNSL depends on the representation of the input. In particular, while the bulk of past theoretical work on the topic assumed the use of the so-called non-zero representation, here we prove that if an additive representation can be used instead then BNSL becomes fixed-parameter tractable even under significantly milder restrictions to the superstructure, notably when parameterized by the treewidth alone. Last but not least, we show how our results can be extended to the closely related problem of Polytree Learning.