报告时间：2019.10.28  16:00-17:00  

报告地点：宝山校区机自大楼702B

报告题目：Distributed Online Convex Optimization with Time-Varying Coupled Inequality Constraints

报告人：易新蕾, KTH Royal Institute of Technology

邀请人： Prof. Xiaoqiang Ren

摘要：In this talk, we consider the problem of distributed online convex optimization with time-varying coupled inequality constraints. This problem can be defined as a repeated game between a group of learners and an adversary. The learners attempt to minimize a sequence of global loss functions and at the same time satisfy a sequence of coupled constraint functions. The global loss and the coupled constraint functions are the sum of local convex loss and constraint functions, respectively, which are generated by the adversary. A distributed online primal-dual gradient descent algorithm is proposed to solve this problem. Without assuming Slater's condition, we prove that the algorithm achieves sublinear dynamic regret and constraint violation if the accumulated variation of the optimal sequence also grows sublinearly. Assuming Slater's condition, we show that the algorithm achieves smaller bounds on the constraint violation. In addition, smaller bounds on the static regret are achieved when the objective function is strongly convex.

报告人简介：Xinlei Yi received the B.S. degree in mathematics from China University of Geoscience, Wuhan, China and M.S.  degree in mathematics from Fudan University, Shanghai, China, in 2011 and 2014,respectively. He is currently pursuing the Ph.D. degree in automatic control at KTH Royal Institute of Technology, Stockholm, Sweden. His current research interests include online optimization, distributed optimization, and event-triggered control.