A gradient-based algorithm for computing Wasserstein barycenter on symmetric cones
太阳成集团tyc411(中国)有限公司-百度百科九十周年院庆系列活动之七十三
题目:A gradient-based algorithm for computing Wasserstein barycenter on symmetric cones
报告人: Sangho Kum教授 (韩国忠北国立大学)
时间:2018年9月22日13:30-14:30
地点:工商楼200-9
摘要: For probability measures with finite second moment, the Monge-Kantorovich problem is important. The distance dW, the square root of the minimum, on the set of probability measures with finite second moment is called the L2-Wasserstein metric. This work is concerned with the least squares problem for the Wasserstein distance dW. The main purpose is to develop gradient-based algorithms for computing the Wasserstein barycenter of Gaussian measures and to provide its convergence analysis especially. Numerical tests will be reported. Moreover, we extend the results into the general setting of symmetric cones for providing a unifying scheme to deal with the problem.
联系人:李冲(cli@zju.edu.cn)