Transfer Reinforcement Learning for Differing Action Spaces via Q-Network Representations.
Nathan BeckAbhiramon RajasekharanHieu TranPublished in: CoRR (2022)
Keyphrases
- action space
- reinforcement learning
- continuous state
- markov decision processes
- state and action spaces
- state space
- function approximators
- real valued
- transfer learning
- function approximation
- action selection
- continuous state spaces
- state action
- model free
- stochastic processes
- multi agent
- optimal policy
- network structure
- temporal difference
- machine learning
- skill learning
- conditional random fields
- graphical models
- supervised learning
- dynamic programming
- decision making
- learning algorithm