深度学习中的无监督学习方法综述-计算机系统应用.PDFVIP

2016 年 第 25 卷 第 8 期 计 算 机 系 统 应 用 深度学习中的无监督学习方法综述① 1 2 2 2 殷瑞刚 , 魏 帅 , 李 晗 , 于 洪 1(解放军第 181 医院, 桂林 541002) 2( 国家数字交换系统工程技术研究中心, 郑州 450002) 摘 要: 从 2006 年开始, 深度神经网络在图像/语音识别、自动驾驶等大数据处理和人工智能领域中都取得了巨 大成功, 其中无监督学习方法作为深度神经网络中的预训练方法为深度神经网络的成功起到了非常重要的作用. 为此, 对深度学习中的无监督学习方法进行了介绍和分析, 主要总结了两类常用的无监督学习方法, 即确定型的 自编码方法和基于概率型受限玻尔兹曼机的对比散度等学习方法, 并介绍了这两类方法在深度学习系统中的应 用, 最后对无监督学习面临的问题和挑战进行了总结和展望. 关键词: 自编码; 受限玻尔兹曼机; 无监督学习; 深度学习; 神经网络 Introduction of Unsupervised Learning Methods in Deep Learning 1 2 2 2 YIN Rui-Gang , WEI Shuai , LI Han , YU Hong 1(Chinese People’s Liberation Army 181th Hospital, Guilin 541002, China) 2(National Digital Switching System Engineering Technological RD Center, Zhengzhou 450002, China) Abstract: Since 2006, Deep Neural Network has achieved huge access in the area of Big Data Processing and Artificial Intelligence, such as image/video discriminations and autopilot. And unsupervised learning methods as the methods getting success in the depth neural network pre training play an important role in deep learning. So, this paper attempts to make a brief introduction and analysis of unsupervised learning methods in deep learning, mainly includs two types, Auto-Encoders based on determination theory and Contrastive Divergence for Restrict Boltzmann Machine based on probability theory. Secondly, the applications of the two methods in Deep Learning are introduced. At last a brief summary and prospect of the challenges faced by unsupervised learning methods in Deep Neural Networks are made. Key words: auto-encoders; restrict boltzmann machine; unsupervised learning; deep learning; neural network 20 世纪八十年代 Hopfield 将能量函数引入到神经 首先从像素中抽象出线条, 再抽象圆形, 再抽象出轮