ICML 2023 SPIGM & LLW Workshop paper: 'The Local Inconsistency Resolution Algorithm'
Developed the theory of Probabilistic Dependency Graphs (PDGs), unifying Bayesian Networks, Factor Graphs, causal models, and Dempster-Shafer belief functions
Showed that standard loss functions correspond to the inconsistency of appropriately constructed PDGs
Designed inference algorithms for PDGs and revealed a deep algorithmic equivalence between inference and inconsistency quantification