RBM学习方法对比  

RBM learning method comparison

作　　者：陆萍[1] 陈志峰[1] 施连敏[1] 

机构地区：[1]苏州经贸职业技术学院信息系,江苏苏州215009

出　　处：《计算机时代》2014年第11期10-13,共4页Computer Era

基　　金：江苏省高校"青蓝工程"优秀青年教师项目;2012江苏省高校产业化推进项目(JHB2012-79);苏州市科技支撑项目(SS201227)

摘　　要：随着深度学习在模型、算法与理论上的突破性进展,以玻尔兹曼机为基础的各类深度模型近年来在目标识别与自然语言处理等诸多领域得到广泛应用。概述了玻尔兹曼机的相关概念,分析了受限玻尔兹曼机模型所具有的优势。对RBM中的学习方法进行了详细的描述,对应用最为广泛的受限玻尔兹曼机的几种典型学习算法进行了对比,并指出学习算法的研究在未来仍将是深度学习中的一项核心问题。With the deep learning on the breakthrough of models, algorithms and theory studies, models based on Boltzmann machine have been used in many areas in recent years, such as target recognition and natural language processing. The concept of Boltzmann machine is presented. The restricted Boltzmann machine＆#39;s advantage is also pointed out. In this paper, the learning method of RBM is described in detail and some typical learning algorithms widely used are compared. The study on learning algorithms will still be a core issue in deep learning area.

关 键 词：受限玻尔兹曼机 深度模型 隐藏单元 学习方法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]