Research map / Optimization / Nonconvex and smooth optimization
A Near-Optimal Algorithm for Decentralized Convex-Concave Finite-Sum Minimax Optimization
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes DIVERSE, a decentralized variance-reduced optimistic gradient method with stochastic mini-batch sizes for convex-concave finite-sum minimax problems over networks, claiming near-optimal complexity.
Problem
Distributed convex-concave finite-sum minimax optimization over a network of agents, where prior methods' convergence depends on local rather than global smoothness parameters.
Approach
A decentralized variance-reduced optimistic gradient method with stochastic mini-batch sizes whose linear rate depends on global smoothness parameters, yielding sharper computation and communication complexity than prior decentralized methods.
Key results
- For strongly-convex-strongly-concave objectives, DIVERSE achieves a linear convergence rate depending on global smoothness parameters, yielding sharper computation and communication complexity per the abstract.
Experimental setup
Metrics: linear convergence rate computation complexity communication complexity
Closest papers by meaning
- Decentralized Convex Finite-Sum Optimization with Better Dependence on Condition Numbers (ICML 2024 · similarity 0.93)
- Jointly Improving the Sample and Communication Complexities in Decentralized Stochastic Minimax Optimization (AAAI 2024 · similarity 0.92)
- Near-Optimal Distributed Minimax Optimization under the Second-Order Similarity (NeurIPS 2024 · similarity 0.91)
- Optimal and Practical Algorithms for Smooth and Strongly Convex Decentralized Optimization (NeurIPS 2020 · similarity 0.90)
- Lower Bounds and Optimal Algorithms for Non-Smooth Convex Decentralized Optimization over Time-Varying Networks (NeurIPS 2024 · similarity 0.89)
- Shuffling Gradient-Based Methods for Nonconvex-Concave Minimax Optimization (NeurIPS 2024 · similarity 0.88)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.