Title：         Algorithms for Group Sparsity with Overlap and Beyond

Speaker：      Weiguo Gao（Fudan University）

Time&Venue：  2016.06.04  15:30pm  N402

Abstract:       We discuss the algorithms for group sparsity with overlap in which matrix-vector multiplications can be carried out efficiently. An accelerating technique to improve the performance of existing algorithms is proposed by solving a similar sub-problem. We further show that the new scheme can be formulated in the inner-outer regime by exploring the sparse structure. This can also be viewed from the dimensional reduction viewpoint. Convergence for ADMM (as an example) is guaranteed by a rigorous proof. Numerical experiments demonstrate the efficiency of the technique.