Koopman Q-learning: Offline Reinforcement Learning via Symmetries of Dynamics.
Matthias WeissenbacherSamarth SinhaAnimesh GargYoshinobu KawaharaPublished in: ICML (2022)
Keyphrases
- reinforcement learning
- function approximation
- reinforcement learning algorithms
- state space
- model free
- learning algorithm
- action selection
- optimal policy
- multi agent
- reinforcement learning methods
- dynamical systems
- markov decision processes
- real time
- dynamic model
- optimal control
- state action space
- temporal difference
- relational reinforcement learning
- eligibility traces
- continuous state and action spaces
- dynamic programming
- stochastic approximation
- machine learning
- hierarchical reinforcement learning
- temporal difference learning
- state action
- function approximators
- control problems
- learning classifier systems
- multi agent reinforcement learning
- learning problems
- action space
- policy iteration
- learning tasks
- continuous state
- policy search
- transfer learning
- agent learns
- cooperative
- reward function