Topic
structure learning
Information Lattice Learning Framework Enables Interpretable Structure Learning for Probabilistic Graphical Models
A new paper by Haizi Yu and Lav R. Varshney introduces Information Lattice Learning (ILL) as a method for interpretable structure learning of probabilistic graphical models. ILL learns rules by alternately projecting signals onto partition lattices and lifting selected rules back, with applications in constraint-based factor graphs and maximum entropy modeling.
ArXiv Paper Introduces KG-SoftMAP: Bayesian Network Learning from Sparse Data Using Knowledge Graph Priors
KG-SoftMAP is a new method for Bayesian network structure learning from sparse discrete data, using a weighted knowledge graph as a prior. On synthetic benchmarks, it achieved Directed-F1 scores up to 0.97 at higher observation rates, while data-only learners stayed near zero. On real educational datasets, it matched logistic regression within 0.03 F1_FAIL while providing an interpretable concept graph.