Research and Development on Inference Technique in Probabilistic Graphical Models
Liu Jian-we · 2015
In recent years,probabilistic graphical models have become the focus of the research in uncertainty inference,because of their bright prospect for the application in artificial intelligence,machine learning,computer vision and so forth.According to different network structures and query questions,the inference algorithms of probabilistic graphical models were summarized in a systematic way.First,exact and approximate inference algorithms for solving the probability queries in Bayesian network and Markov network were discussed,including variable elimination algorithms,conditioning algorithms,clique tree algorithms,variational inference algorithms and sampling algorithms.The common algorithms for solving MAP queries were also introduced.Then the inference algorithms in hybrid networks were described respectively for continuous or hybrid cases.In addition,this work analyzed the exact and approximate inference in temporal networks,and described inference in continuous or hybrid cases for temporal networks.Finally,this work raised some questions that the inference algorithms of probabilistic graphical models are facing with and discussed their development in the future.