講座題目：Sensitivity-Based Robust Learning Algorithm for Stacked Autoencode

報 告 人：Prof. Daniel S Yeung (Hong Kong Polytechnic University)

時　 間：2018年10月29日(周一)10:00

地　 點：中關村校區信息科學實驗樓205室

主辦單位：研究生院、信息與電子學院

報名方式：登錄北京理工大學微信企業號---第二課堂---課程報名中選擇“百家大講堂第125期：Sensitivity-Based Robust Learning Algorithm for Stacked Autoencode”

【主講人簡介】

Daniel S. Yeung (Ph.D., M.Sc., M.B.A., M.S., M.A., B.A.) is a Past President of the IEEE Systems, Man and Cybernetics (SMC) Society and a Fellow of IEEE. He received the Ph.D. degree in applied mathematics from Case Western Reserve University. In the past, he has worked as an Assistant Professor of Mathematics and Computer Science at Rochester Institute of Technology, as a Research Scientist in the General Electric Corporate Research Center, and as a System Integration Engineer at TRW, all in the United States. He was the chairman of the Department of Computing, The Hong Kong Polytechnic University, Hong Kong, and a Chair Professor from 1999 to 2006. He had also served as a visiting Professor in the School of Computer Science and Engineering, South China University of Technology, Guangzhou, China.

His current research interests include machine learning, deep learning, neural-network sensitivity analysis, and big data analytic. He is the Co-EiC of the international Journal on Machine Learning and Cybernetics (Springer). He was the Chairman of IEEE Hong Kong Computer Chapter (91and 92), an associate editor for both IEEE Transactions on Neural Networks and IEEE Transactions on SMC (Part B), and for the International Journal on Wavelet and Multiresolution Processing. He has served as the President (2008 and 2009), President-Elect (2007), a member of the Board of Governors, Vice President for Technical Activities, and Vice President for Long Range Planning and Finance for the IEEE SMC Society. He co-founded and served as a General Co-Chair since 2002 for the International Conference on Machine Learning and Cybernetics held annually in China. He also served as a General Co-Chair (Technical Program) of the 2006 International Conference on Pattern Recognition, and 2012, 2013 and 2015 International Conference on Systems, Man and Cybernetics. He is also the founding Chairman of the IEEE SMC Hong Kong Chapter. 

【講座摘要】

Although deep learning has achieved excellent performance in many applications, some studies indicate that deep learning algorithms are vulnerable in an adversarial environment. A small distortion on a sample leads to misclassification easily. Until now, the vulnerability issue of stacked autoencoder, which is one of the most popular deep learning algorithms, has not been investigated. In this talk, a robust learning algorithm which minimizes both its error and sensitivity is proposed for stacked autoencoder. The sensitivity is defined as the change of the output due to a small fluctuation on the input. As the proposed algorithm considers not only accuracy but also stability, a more robust stacked autoencoder against perturbed input is expected. The performance of our methods is then evaluated and compared with conventional stacked autoencoder and denoising autoencoder experimentally in terms of accuracy, robustness and time complexity. Moreover, the experimental results also suggest that the proposed learning method is more robust than others when a training set is contaminated.