应数学系马坚伟教授邀请，美国UCLA著名学者Wotao Yin教授，访问我系一周。将在002全讯白菜网做题为稀疏优化理论与算法的四个系列报告。欢迎大家参加。
专家简介：
印卧涛 (Wotao Yin),美国UCLA数学系教授。1979年生于南京，本科毕业于南京大学数学系，2006年获Columbia University运筹学博士学位。2006年-2013年7月，Rice University计算与应用数学系助理教授、副教授。
从事压缩感知和矩阵恢复理论算法研究，已在SIAM Journal 发表论文20多篇，IEEE PAMI等发表IEEE系列论文多篇，近五年论文被引用约5000余次(Google scholar)，单篇最高被引703次，单篇被引用过百次的论文共14篇。开发了大量程序代码可供同行下载使用，比较知名的有：LMaFit: low-rank matrix completion (2012)，YALL1 and YALL1-Group: dual-ADMM based algorithms for l1 and group/joint-l1 (2011)，RecPF: compressive sensing from incomplete Fourier samples, ADMM (2010)，FTVd: total variation color image denoising and deblurring, ADMM (2010)，FPC: l1 solver by iterative shrinkage and continuation, with Barzilai-Borwein step (2008)。
题目1：Distributed and decentralized sparse optimization, Parts 1 & 2
时间：2013.12.24 14:00-16:00
地点：一区活动中心326
内容：Modern datasets usually have a large number of features or training samples, and they are usually stored in a distributed manner. These two lectures will review the basics of parallel computing, existing sparse optimization algorithms, and introduces ones tailored for very large scale sparse optimization problems such as LASSO, basis pursuit, exchange problem and other problems with terabytes of data through distributed and/or decentralized computation.
题目2：New sparse regularization evolving l1 subgradient, Parts 1 & 2
时间：2013.12.26 14:00-16:00
地点：一区活动中心326
内容：We introduce new spare regularization approaches based on evolving the subgradient vector p of the l1-norm. It is closely related to LASSO, inverse scale space, Bregman regularization, and linearized Bregman. We show that the new approaches give better solutions than LASSO. The solutions are sparser and fit data better. We show that on the so-called solution path, there will exist a time t where the solution will have the same signs as the true signal and a so-called oracle property is satisfied. Both theoretical guarantees and computational results will be given. This is joint work with Ming Yan, Yuan Yao, Stanley Osher, and others.
