加入收藏夹
联系我们
关于本站
个人主页
西电导航
西安电子科技大学
    当前位置:首页>>学术讲座
Interpretable Convolutional Neural Networks (CNNs) via Feedforward Design
时间:2019-06-13 08:33    点击:   所属单位:人工智能学院
讲座名称 Interpretable Convolutional Neural Networks (CNNs) via Feedforward Design
讲座时间 2019-06-14 10:00:00
讲座地点 主楼II-221
讲座人 C.-C Jay Kuo
讲座人介绍
Prof. C.-C. Jay Kuo of University of Southern California  received his Ph.D. degree from the Massachusetts Institute of Technology in 1987. He is now with the University of Southern California (USC) as Director of the Media Communications Laboratory and Distinguished Professor of Electrical Engineering and Computer Science. His research interests are in the areas of media processing, compression and understanding. Dr. Kuo was the Editor-in-Chief for the IEEE Trans. on Information Forensics and Security in 2012-2014. Dr. Kuo is a Fellow of AAAS, IEEE and SPIE. He has guided 150 students to their Ph.D. degrees and supervised 29 postdoctoral research fellows. Dr. Kuo is a co-author of 280 journal papers, 920 conference papers and 14 books. Dr. Kuo received the 2016 IEEE Computer Society Taylor L. Booth Education Award, the 2016 IEEE Circuits and Systems Society John Choma Education Award, the 2016 IS&T Raymond C. Bowman Award, the 2017 IEEE Leon K. Kirchmayer Graduate Teaching Award, the 2017 IEEE Signal Processing Society Education Award, and the 2019 IEEE Computer Society Edward J. McCluskey Technical Achievement Award.

讲座内容
Given a convolutional neural network (CNN) architecture, its network parameters are determined by backpropagation (BP) nowadays. The underlying mechanism remains to be a black-box after a large amount of theoretical investigation. In this talk, I describe a new interpretable and feedforward (FF) design with the LeNet-5 as an example. The FF-trained CNN is a data-centric approach that derives network parameters based on training data statistics layer by layer in one pass. To build the convolutional layers, we develop a new signal transform, called the Saab (Subspace approximation with adjusted bias) transform. The bias in filter weights is chosen to annihilate nonlinearity of the activation function. To build the fully-connected (FC) layers, we adopt a label-guided linear least squared regression (LSR) method. The classification performances of BP- and FF-trained CNNs on the MNIST and the CIFAR-10 datasets are compared. The computational complexity of the FF design is significantly lower than the BP design and, therefore, the FF-trained CNN is ideal for mobile/edge computing. We also comment on the relationship between BP and FF designs by examining the cross-entropy values at nodes of intermediate layers.

转载请注明出处:西安电子科技大学学术信息网
如果您有学术信息或学术动态,欢迎投稿。我们将在第一时间确认并收录,投稿邮箱: meeting@xidian.edu.cn
Copyright © 2011-2019 西安电子科技大学 
开发维护:电子工程学院网络信息中心  管理员:meeting@xidian.edu.cn 站长统计: