»ã±¨Ö÷Ì⣺A block symmetric Gauss-Seidel decomposition theorem for convex quadratic programming and its applications (¿é¶Ô³Æ¸ß˹-ÈûµÃ·Ö»¯¶¨ÀíÔÚ͹¶þ´Î¹æ»®ÖеÄÀûÓÃ)

»ã±¨ÈË£ºKim-Chuan Toh ½ÌÊÚ£¨Ð¼ÓƹúÁ¢´óѧÊýѧϵ£©

»ã±¨¹¦·ò£º2020Äê9ÔÂ21ÈÕ£¨ÖÜÒ»£© 14:00-16:00

²Î»á·½Ê½£ºÌÚѶ »áÒé

»áÒéID£º302 840 008

»áÒéÃÜÂ룺200921

»áÒéÁ´½Ó£ºhttps://meeting.tencent.com/s/Xbk2rAnqeKOE

Ö÷°ì²¿ÃÅ£ºÐ±¦GGÔ˳ïÓëÓÅ»¯Ê¢¿ª³¢ÊÔÊÒ-¹ú¼Ê¿ÆÑкÏ×÷ƽ̨¡¢ÉϺ£ÊÐÔ˳ïѧ»á¡¢Ð±¦GGÀíѧԺÊýѧϵ

»ã±¨ÌáÒª£ºFor a multi-block convex composite quadratic programming (CCQP) with an additional nonsmooth term in the first block, we present a block symmetric Gauss-Seidel (sGS) decomposition theorem, which states that each cycle of the block sGS method is equivalent to solving the CCQP with an additional proximal term constructed from the sGS decomposition of the quadratic term. As a basic building block, the sGS decomposition theorem has played a key role in various recently developed algorithms such as the inexact proximal ALM/ADMM for linearly constrained multi-block convex composite conic programming. We demonstrate how our sGS-based ADMM can be applied to solve doubly nonnegative semidefinite programming and Wasserstein barycenter problems.

Ó­½ÓÀÏʦ¡¢Ñ§Éú²ÎÓ룡