Bayesian Optimal Experimental Design for Inferring Causal Structure

Inferring the causal structure of a system typically requires interventional data, rather than just observational data. Since experiments can be costly, it is preferable to select interventions that yield maximum amount information about system. We propose novel Bayesian method for optimal experimental design by sequentially selecting minimize expected posterior entropy as rapidly possible. A key feature implemented computing simple summaries current posterior, avoiding computationally burdensome task repeatedly performing inference on hypothetical future datasets drawn from predictive. After deriving in general setting, we apply problem inferring networks. present series simulation studies, which find proposed performs favorably compared existing alternative methods. Finally, real data two gene regulatory