机构地区: 华中理工大学电子与信息工程系
出 处: 《华中理工大学学报》 1993年第6期1-5,共5页
摘 要: 随着VLSI技术的迅猛发展,对神经网络与隐马尔柯夫模型(HMM’s)之问的关系研究已成为信息处理领域的一个重要的研究方向.在分析高阶神经网络和二阶HMM’s的结构及算法的基础上,提出了这两种模型的统一数学模型,从而为这两种模型的systolic设计奠定了基础. With the rapid progress made in VLSI technology, investigations on the relationship between neural networks and hidden Markov models (HMM's) have become an important research direction in the field of information processing. In this paper, the relationship between the high order multilayer neural network (HMNN) and second order HMM is studied. The architecture of recurrent HMNN is presented and the retrieving and learning phases of recurrent HMNN are discussed. The structure and algorithm of the second order HMM are described. Based on the close algorithmic similarity in the learning technique, a unified formulation for the two models is proposed.