MALib: A Parallel Framework for Population-based Multi-agent Reinforcement Learning 论文笔记

文章目录

• 前言
• Parallel Programming Abstractions for PB-MARL
• • Centralized Task Dispatching Model
  • • Defining the Task Graph
    • Decoupling the Task-Data Flow
  • Actor-Evaluator-Learner Model
  • Abstractions for MARL Training Paradigms
  • • Independent Learning
    • Centralized Learning
    • Asynchronous Learning
    • Synchronous Learning