Koopman Q-learning: Offline Reinforcement Learning via Symmetries of Dynamics.
Matthias WeissenbacherSamarth SinhaAnimesh GargYoshinobu KawaharaPublished in: CoRR (2021)
Keyphrases
- reinforcement learning
- function approximation
- reinforcement learning algorithms
- model free
- state space
- reinforcement learning methods
- temporal difference learning
- control problems
- optimal policy
- real time
- learning algorithm
- multi agent reinforcement learning
- state action space
- dynamical systems
- markov decision processes
- temporal difference
- dynamic programming
- machine learning
- rl algorithms
- multi agent
- action selection
- dynamic model
- eligibility traces
- transfer learning
- learning process
- policy search
- optimal control
- stochastic approximation
- continuous state spaces
- reward shaping
- neural network
- cooperative
- policy iteration
- supervised learning
- feature points