A Connection between One-Step RL and Critic Regularization in Reinforcement Learning.
Benjamin EysenbachMatthieu GeistSergey LevineRuslan SalakhutdinovPublished in: ICML (2023)
Keyphrases
- reinforcement learning
- actor critic
- reinforcement learning algorithms
- function approximation
- temporal difference
- model free
- state space
- policy gradient
- learning algorithm
- rl algorithms
- markov decision processes
- approximate dynamic programming
- machine learning
- multi agent
- learning problems
- temporal difference learning
- reinforcement learning methods
- learning agents
- control problems
- neuro fuzzy
- supervised learning
- reinforcement learning problems
- action selection
- gradient method
- optimal policy
- dynamic programming
- function approximators
- policy iteration
- complex domains
- learning classifier systems
- optimal control
- transfer learning
- learning process
- partially observable domains
- direct policy search