Active Learning for Causal Bayesian Network Structure with Non-symmetrical Entropy

Causal knowledge is crucial for facilitating comprehension, diagnosis, prediction, and control in automated reasoning. Active learning in causal Bayesian networks involves interventions by manipulating specific variables, and observing the patterns of change over other variables to derive causal knowledge. In this paper, we propose a new active learning approach that supports interventions with node selection. Our method admits a node selection criterion based on non-symmetrical entropy from the current data and a stop criterion based on structure entropy of the resulting networks. We examine the technical challenges and practical issues involved. Experimental results on a set of benchmark Bayesian networks are promising. The proposed method is potentially useful in many real-life applications where multiple instances are collected as a data set in each active learning step.