文件名称：energy minimization methods in computer vision and pattern recognition
文件大小：42.77MB
文件格式：PDF
更新时间：2017-06-15 15:30:04
能量最小化
Local belief propagation(BP)algorithms[12]have long been known to converge to the correct marginal probabilities for tree-like graphical models. Recently,several researchers have empirically demonstrated that BP algorithms often perform surprisingly well for graphs with loops[4],[5],[10],[11]and turbo decoding [10] . (See[11],however, for examples where BP algorithms fail). It is important to understand when and why BP ,and related algorithms,perform well on graphs with loops.