基于稀疏降噪自编码器的深度置信网络  

Deep belief networks based on sparse denoising auto encoders

作　　者：曾安[1] 张艺楠[1] 潘丹[2] Xiao-Wei Song 

机构地区：[1]广东工业大学计算机学院,广州510006 [2]广东建设职业技术学院现代教育技术中心,广州510440 [3]西蒙弗雷泽大学影像技术实验室

出　　处：《计算机应用》2017年第9期2585-2589,共5页journal of Computer Applications

基　　金：国家自然科学基金资助项目(61300107);广东省自然科学基金资助项目(S2012010010212);广州市科技计划资助项目(201504301341059;201505031501397)~~

摘　　要：传统的深度置信网络(DBN)采用随机初始化受限玻尔兹曼机(RBM)的权值和偏置的方法初始化网络。虽然这在一定程度上克服了由BP算法带来的易陷入局部最优和训练时间长的问题,但随机初始化仍然会导致网络重构和原始输入的较大差别,这使得网络无论在准确率还是学习效率上都无法得到进一步提升。针对以上问题,提出一种基于稀疏降噪自编码器(SDAE)的深度网络模型,其核心是稀疏降噪自编码器对数据的特征提取。首先,训练稀疏降噪自编码;然后,用训练后得到的权值和偏置来初始化深度置信网络;最后,训练深度置信网络。在Poker Hand纸牌游戏数据集和MNIST、USPS手写数据集上测试模型性能,在Poker Hand数据集下,方法的误差率比传统的深度置信网络降低46.4%,准确率和召回率依次提升15.56%和14.12%。实验结果表明,所提方法能有效地改善模型性能。The conventional Deep Belief Network （DBN） often utilizes the method of randomly initializing the weights and bias of Restricted Boltzmann Machine（RBM） to initialize the network. Although it could overcome the problems of local optimality and long training time to some extent, it is still difficult to further achieve higher accuracy and better learning efficiency owing to the huge difference between reconstruction and original input resulting from random initialization. In view of the above-mentioned problem, a kind of DBN model based on Sparse Denoising AutoEncoder （SDAE） was proposed. The advantage of the advocated model was the feature extraction by SDAE. Firstly, SDAE was trained, and then, the obtained weights and bias were utilized to initialize DBN. Finally, DBN was trained. Experiments were performed on card game data set of Poker hand and handwriting data sets of MNIST and USPS to verify the performance of the proposed model. In Poker hand data set, compared with the conventional DBN, the error rate of the proposed model is lowered by 46.4%, the accuracy rate and the recall rate are improved by 15.56% and 14.12% respectively. The results exhibit that the proposed method is superior to other existing methods in recognition performance.

关 键 词：深度置信网络 受限玻尔兹曼机 稀疏降噪自编码器 深度学习 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]