Quotient Normalized Maximum Likelihood Criterion for Learning Bayesian Network Structures

Abstract

We introduce an information theoretic criterion for Bayesian network structure learning which we call quotient normalized maximum likelihood (qNML). In contrast to the closely related factorized normalized maximum likelihood criterion, qNML satisfies the property of score equivalence. It is also decomposable and completely free of adjustable hyperparameters. For practical computations, we identify a remarkably accurate approximation proposed earlier by Szpankowski and Weinberger. Experiments on both simulated and real data demonstrate that the new criterion leads to parsimonious models with good predictive accuracy.

Accepted to AISTATS 2018

Cited by in corpus (8)

• D-VAE: A Variational Autoencoder for Directed Acyclic Graphs
• Causal datasheet: An approximate guide to practically assess Bayesian networks in the real world
• Approximate learning of high dimensional Bayesian network structures via pruning of Candidate Parent Sets
• Evaluating structure learning algorithms with a balanced scoring function
• Greedy structure learning from data that contain systematic missing values
• Estimating Conditional Mutual Information for Discrete-Continuous Mixtures using Multi-Dimensional Adaptive Histograms
• Prequential MDL for Causal Structure Learning with Neural Networks
• Discriminative, Generative and Self-Supervised Approaches for Target-Agnostic Learning