基于遗传算法的HMM参数估计  

Traning of HMM With Genetic Algorithms

作　　者：徐丽[1] 康瑞华[1] 

机构地区：[1]湖北工业大学计算机学院,湖北武汉430068

出　　处：《湖北工业大学学报》2006年第4期68-71,共4页Journal of Hubei University of Technology

摘　　要：隐马尔可夫模型(Hidden Markov model)用于多序列比对研究是生物信息学研究的新领域,其可以通过训练识别同一特征的蛋白质序列.然而,目前的HMM参数估计算法Viterbi算法和Baum-Welch算法,都只能找到局部最优比对,无法找到全局最优比对.针对此算法全局最优问题提出了基于遗传算法的HMM参数估计,与已有的训练算法相比,遗传算法在搜索全局最优时具有突出的优势.Hidden Markov model used in multiple alignment is a new field of bioinformatics. It can be used to distinguish protein sequence with the same characterstics; But the problem is that the two parameter estimation algorithms for HMMs, the Viterbi algorithm and Baum Welch algorithm, are only guaranteed to find a local optimum, sO a novel training method based on genetic algorithm is designed in this paper. Experimental results shows that this algorithm is more superior to those traditional methods in finding a global optimum.

关 键 词：隐马尔可夫模型 遗传算法 多序列比对 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]