报告题目 (Title):A Single-Loop Algorithm for Decentralized Bilevel Optimization(求解去中心化双层优化问题的单循环算法)
报告人 (Speaker):杨俊锋 教授(南京大学)
报告时间 (Time):2025年07月1日(周二) 8:00
报告地点 (Place):校本部GJ303
邀请人(Inviter):周安娃
报告摘要: Bilevel optimization (BO) has gained significant attention in recent years due to its broad applications in machine learning. In this talk, we focus on decentralized BO and proposes a novel single-loop algorithm for solving it with a strongly convex lower-level problem. Our approach is a fully single-loop method that approximates the hypergradient using only two matrix-vector multiplications per iteration. Our algorithm does not require any gradient heterogeneity assumption and achieves the best-known convergence rate for BO algorithms. We also present experimental results on hyperparameter optimization problems using both synthetic and MNIST datasets, which demonstrate the efficiency of our proposed algorithm.