Continuous-time multi-agent systems for distributed minimization and minimax optimization----系统科学研究所

学术报告

Continuous-time multi-agent systems for distributed minimization and minimax optimization
主讲：Dr. Shaofu Yang
举办时间：2016.12.20;9:30am 地点：N226

Abstract: In this talk, I will discuss several continuous multi-agent systems for two classes of distributed optimization problems: distributed minimization and distributed minimax optimization. First, a multi-agent system with a proportional-integral consensus protocol is proposed for distributed minimization subject to general constraints. Under the assumption of local convexity of objective functions, it is proved that all agents with any initial state can reach output consensus at an optimal solution to the given constrained optimization problem. Second, a cooperative-competitive multi-agent system is designed for distributed minimax optimization. All agents in the system are divided into two groups for minimizing and maximizing respectively. The system features competitive inter-group interactions and cooperative intra-group interactions. Under certain conditions, the system is proved to be convergent to a saddle point that solves the minimax optimization problem. As a special case, the system is further applied for distributed minimization, which implies some alternative multi-agent systems for distributed minimization

附件下载：