【15-36期VALSE Webinar活动】 报告嘉宾:杨颖振(UIUC ECE Dept.) 报告时间:2015年12月02日(星期三)晚20:30(北京时间) 报告题目: Learning with L0-Graph: L0-Induced Sparse Subspace Clustering [Slides] 主持人:禹之鼎(CMU ECE Dept.) 报告摘要: Sparse subspace clustering methods, such as Sparse Subspace Clustering (SSC) and L1-graph, are effective in partitioning the data that lie in a union of subspaces. Most of those methods use L1-norm or L2 -norm with thresholding to impose the sparsity of the constructed sparse similarity graph, and certain assumptions, e.g. independence or disjointness, on the subspaces are required to obtain the subspace-sparse representation, which is the key to their success. Such assumptions are not guaranteed to hold in practice and they limit the application of sparse subspace clustering on subspaces with general location. In this paper, we propose a new sparse subspace clustering method named L0-graph. In contrast to the required assumptions on subspaces for most existing sparse subspace clustering methods, it is proved that subspace- sparse representation can be obtained by L0-graph for arbitrary distinct underlying subspaces almost surely under the mild i.i.d. assumption on the data generation. We develop a proximal method to obtain the sub- optimal solution to the optimization problem of L0-graph with proved guarantee of convergence. Moreover, we propose a regularized L0-graph that encourages nearby data to have similar neighbors so that the similarity graph is more aligned within each cluster and the graph connectivity issue is alleviated. Extensive experimental results on various data sets demonstrate the superiority of L0-graph compared to other competing clustering methods, as well as the effectiveness of regularized L0-graph. 参考文献: [1] Yingzhen Yang, Jiashi Feng, Jianchao Yang, and Thomas S. Huang. Learning with L0-Graph: L0-Induced Sparse Subspace Clustering. (Arxiv:1510.08520, 2015) 报告人简介: Yingzhen Yang is a Ph.D. candidate with the Department of Electrical and Computer Engineering, University of Illinois at Urbana-Champaign, under the supervision of Professor Thomas S. Huang. His current research interests include machine learning with focus on statistical learning theory, large-scale probabilistic graphical models, sparse coding, manifold learning and nonparametric methods; computer vision with focus on image classification using deep learning methods and image/video enhancement. He is the recipient of "Lu Zeng Yong" CAD&CG High-Tech Award in China in 2009 and Carnegie Institute of Technology Dean's Tuition Fellowship in 2010. |