题 目：Structured Sparsity, Low Rank and Optimization
报告人： Dr. Andreas Argyriou
（Toyota Technological Institute at Chicago）
时 间: 6月17日（周五） 下午 4：00－5：00
地 点：数学楼210室
摘要：
I will discuss some regularization based methodologies for statistical and machine learning problems. First, I will present a general framework for structured sparsity which can replace ad hoc approaches.It provides a straightforward way of favoring prescribed sparsity patterns, such as orderings, contiguous regions and overlapping groups, among others. I will also discuss regularization approaches involving the trace norm which can be used for multitask learning and low rank problems. To solve nonsmooth convex problems like the above, first-order methods from optimization can be used. I will present a general view and variations on such methods based on fixed points and successive approximations.
欢迎师生们踊跃参加！