Parameter Sharing is Surprisingly Useful for Multi-Agent Deep Reinforcement Learning.
Justin K. TerryNathaniel GrammelAnanth HariLuis SantosBenjamin BlackDinesh ManochaPublished in: CoRR (2020)
Keyphrases
- reinforcement learning
- multi agent
- function approximation
- multi agent environments
- state space
- intelligent agents
- cooperative
- multi agent systems
- parameter values
- reinforcement learning agents
- learning agents
- reinforcement learning algorithms
- information sharing
- robotic control
- markov decision processes
- knowledge sharing
- learning algorithm
- multiagent systems
- single agent
- temporal difference
- dynamic programming
- counter intuitive
- multi agent reinforcement learning
- supervised learning
- action selection
- model free
- coalition formation
- data sharing
- autonomous agents
- dynamical systems
- optimal policy
- action space
- reinforcement learning methods
- learning problems
- software agents
- learning process
- multi agent coordination
- agent based simulations