Dr. Hyunwoong Chang

Title:  Order-based Structure Learning without Score Equivalence

Date: Thursday, May 30th, 2024
Time: 1:30PM (PDT)
Location: ASB 10900

Abstract: Structure learning of directed acyclic graph (DAG) models is the task of discovering the underlying DAG structure that represents the conditional independence relations among variables in a given observational data. In this talk, we propose an empirical Bayes formulation of the structure learning, where the prior specification assumes that all node variables have the same error variance, an assumption known to ensure the identifiability of the underlying causal DAG. To facilitate efficient posterior computation, we approximate the posterior probability of each ordering by that of a best DAG model, which naturally leads to an order-based Markov chain Monte Carlo (MCMC) algorithm. Strong selection consistency for our model in high-dimensional settings is proved under a condition that allows heterogeneous error variances, and the mixing behavior of our sampler is theoretically investigated. We demonstrate that our method outperforms other state-of-the-art algorithms under various simulation settings, and we provide a single-cell real-data study to illustrate the practical advantages of the proposed method.