本章目录

22 More variational inference 767
22.1 Introduction 767
22.2 Loopy belief propagation: algorithmic issues 767
22.2.1 A brief history 767
22.2.2 LBP on pairwise models 768
22.2.3 LBP on a factor graph 769
22.2.4 Convergence 771
22.2.5 Accuracy of LBP 774
22.2.6 Other speedup tricks for LBP * 775
22.3 Loopy belief propagation: theoretical issues * 776
22.3.1 UGMs represented in exponential family form 776
22.3.2 The marginal polytope 777
22.3.3 Exact inference as a variational optimization problem 778
22.3.4 Mean field as a variational optimization problem 779
22.3.5 LBP as a variational optimization problem 779
22.3.6 Loopy BP vs mean field 783
22.4 Extensions of belief propagation * 783
22.4.1 Generalized belief propagation 783
22.4.2 Convex belief propagation 785
22.5 Expectation propagation 787
22.5.1 EP as a variational inference problem 788
22.5.2 Optimizing the EP objective using moment matching 789
22.5.3 EP for the clutter problem 791
22.5.4 LBP is a special case of EP 792
22.5.5 Ranking players using TrueSkill 793
22.5.6 Other applications of EP 799
22.6 MAP state estimation 799
22.6.1 Linear programming relaxation 799
22.6.2 Max-product belief propagation 800
22.6.3 Graphcuts 801
22.6.4 Experimental comparison of graphcuts and BP 804
22.6.5 Dual decomposition 806

github下载链接：https://github.com/916718212/Machine-Learning-A-Probabilistic-Perspective-.git