时间：3月19日星期二下午14:00
地点：浙大紫金港校区CAD&CG国家重点实验室402室
报告题目：Matrix Completion for Visual Applications and Recommender System
报告人：胡尧
主持人：张德兵

Abstract：Recovering a large matrix from a small subset of its entries is a challenging problem arising in many real applications, such as image inpainting and recommender systems. Many existing approaches formulate this problem as a general low rank matrix approximation problem. Since the rank operator is non-convex and discontinuous, most of the recent theoretical studies use the nuclear norm as a convex relaxation. One major limitation of the existing approaches based on nuclear norm minimization is that the rank may not be well approximated in practice. We introduce a better approximation to the rank of matrices by Truncated Nuclear Norm for the completion of missing information in the visual applications. Empirical study show encouraging results of the proposed algorithm for image inpainting. Furthermore, we also propose to accelerate the traditional singular value threhsolding algorithm for large scale matrix completion problem for recommender system. Theoretical analysis and experimental results both show that our proposed algorithm can greatly accelerate the convergence.

Bio：Yao Hu received the BS degree in Math and Applied Mathematics from Zhejiang University, China, in 2010. He is a currently a Ph.D candidate in Computer Science at Zhejiang University. His research interests include machine learning, computer vision and data mining.