























The paper presents Entropy Semiring Forwardbackward algorithm (ESRFB) and its application for memory efficient computation of the subsequence constrained entropy and state sequence entropy of a Hidden Markov Model (HMM) when an observation sequence is given. ESRFB is based on forward-backward recursion over the entropy semiring, having the lower memory requirement than the algorithm developed by Mann and MacCallum, with the same time complexity. Furthermore, when it is used with forward pass only, it is applicable for the computation of HMM entropy for a given observation sequence, with the same time and memory complexity as the previously developed algorithm by Hernando et al
此内容由惯性聚合(RSS阅读器)自动聚合整理,仅供阅读参考。 原文来自 — 版权归原作者所有。