REBEL: Reinforcement Learning via Regressing Relative Rewards.
Zhaolin GaoJonathan D. ChangWenhao ZhanOwen OertellGokul SwamyKianté BrantleyThorsten JoachimsJ. Andrew BagnellJason D. LeeWen SunPublished in: CoRR (2024)
Keyphrases
- reinforcement learning
- markov decision processes
- learning algorithm
- function approximation
- state space
- model free
- reward function
- optimal policy
- machine learning
- reinforcement learning methods
- reinforcement learning algorithms
- reward shaping
- temporal difference learning
- multi agent
- dynamic programming
- learning problems
- temporal difference
- action selection
- website
- robotic control
- multi agent reinforcement learning
- hidden state
- partial observability
- database
- control policy
- action space
- policy iteration
- markov decision process
- robot control
- partially observable
- transfer learning
- real time