Stochastic margin-based structure learning of Bayesian network classifiers

► Stochastic margin-based structure learning of Bayesian network classifiers (BNC). ► We introduce the maximum margin criterion for discriminative structure learning of BNC. ► We empirically show its advantages using deterministic/stochastic search heuristics. ► We compare generative/discriminative parameter learning on generative/discriminative structured BNC. ► Missing features can be handled by discriminatively optimized BNC.