报告人：张敏 副教授 (广州大学)

时间：2025年03月21日 16:00-

地点：数统学院LD416

摘要：This paper studies a class of conic linkage problems, a generalization of the classical linkage problem involving subspaces, which frequently arise in applications such as multistage stochastic variational inequalities and conic complementarity problems. We propose a conic progressive decoupling algorithm (CPDA) that extends the progressive decoupling approach to conic constraints. By leveraging splitting techniques rather than Spingarn’s partial inverse framework, CPDA iteratively solves a generalized equation involving the monotone operator and updates projections onto the conic set and its dual. We establish the convergence of CPDA and analyze its rate of convergence under specific structural conditions. Numerical experiments on two-stage stochastic complementarity problems and multistage stochastic programming demonstrate the effectiveness of our approach compared to existing decomposition methods.

邀请人: 蒋杰

欢迎广大师生积极参与！