Structure Learning of Large Scale Bayesian Network

We improve the structure learning approach from several aspects to learn huge Bayesian network and propose network merging methods to get better result. This approach is applied to build mRNA-miRNA-cancer network by using dataset whose samples have both mRNAs and miRNAs expression data. We evaluate the learning approach and compare merging methods through experiments and evaluate the network we have learned. Experiments show that the gene interact relationship and even causal relationship can be revealed to get better understanding of the way they interact.