Towards Efficient Local Causal Structure Learning

Local causal structure learning aims to discover and distinguish direct causes (parents) effects (children) of a variable interest from data. While emerging successes have been made, existing methods need search large space target T. To tackle this issue, we propose novel Efficient Causal Structure algorithm, named ELCS. Specifically, first the concept N-structures, then design an efficient Markov Blanket (MB) discovery subroutine integrate MB with N-structures learn T simultaneously With proposed subroutine, ELCS starts variable, sequentially finds MBs variables connected constructs local structures over until distinguished. Using eight Bayesian networks extensive experiments validated that achieves better accuracy efficiency than state-of-the-art algorithms.