A MCMC Algorithm for Improved Bayesian Network Structure Learning

The Bayesian network is a probabilistic graph model based on causal modeling and uncertainty reasoning. It gives the probability relationship between variables so that the posterior probability of the desired result can be calculated. In the fields of machine learning and artificial intelligence, it is commonly employed. The structural learning of Bayesian networks is the focus and complexity of study since the variables can build multiple network topologies and so form complex models. The application of an enhanced MCMC approach for network structure search is investigated in this research. A multi-step proposed distribution based on the Metropolis-Hastings method decreases computational cost, optimizes the solution process, and enhances the MCMC chain convergence speed. The approach is also used to solve a Bayesian network structure learning problem. The results of the experiments demonstrate.