《受限波尔兹曼机简介》.pdf.pdf

受限波尔兹曼机简介张春霞 1,姬楠楠 1,王冠伟 2 1西安交通大学数学与统计学院,西安710049 2西安交通大学机械学院,西安710049 摘要:受限波尔兹曼机(Restricted Boltzmann Machines, RBM)是一类具有两层结构、对称连接且无自反馈的随机神经网络模型,层间全连接,层内无连接。近年来,随着RBM的快速学****算法-对比散度的出现,机器学****界掀起了研究RBM理论及应用的热潮。实践表明, RBM是一种有效的特征提取方法,用于初始化前馈神经网络可明显提高泛化能力,堆叠多个RBM组成的深度信念网络能提取更抽象的特征。鉴于RBM的优点及其广泛应用,本文对RBM的基本模型、学****算法、参数设置、评估方法、变形算法等进行了详细介绍,最后探讨了RBM在未来值得研究的方向。关键词:机器学****深度学****受限波尔兹曼机;对比散度; Gibbs采样. 中图分类号:TP181(自动推理、机器学****O235(模式识别理论) Introduction of Restricted Boltzmann Machines ZHANG Chun-Xia 1, JI Nan-Nan 1, WANG Guan-Wei 2 1School of Mathematics and Statistics, Xi’an Jiaotong University, Xi’an 710049 2School of Mechanical Engineering, Xi’an Jiaotong University, Xi’an 710049 Abstract:A Restricted Boltzmann Machine (RBM) is a particular type of random work model which has two-layer architecture, symmetric connections and no self-feedback. The two layers in an RBM are fully connected but there are no connections within the same layer. Recently, with the advent of a fast learning algorithm for RBM (., contrastive divergence), the machine munity set o? a surge to study the theory and applications of RBM since it has many advantages. For example, RBM is an e?ective method to detect features. When a feed-forward work is initialized with an RBM, its generalization capability can be signi?cantly improved. A deep posed of several RBMs can detect more abstract features. Due to the advantages and wide applications of RBM, this paper attempts to provide a started guide for novice. It presents a detailed introduction of basic RBM model, its representative learning algorithm, parametric settings, evaluation methods, its variants and etc. Finally, some research directions of RBM that are deserved to be further studied are discussed. Key words:Machine learning; Deep learning; Restricted Boltzmann machine; Contrastive 基金项目:高等学校博士学科点专项科研基金(20100201120048),数学天元青年基金(11126277),国家自然科学基金(11201367),中央高校基本科研业务费专项基金. 作者简介:张春霞(1980-),女,讲师,主要研究方向:模式识别、(1985-),女,博士生,主要研究方向:深度学****1979-