State Space Decomposition and Subgoal Creation for Transfer in Deep Reinforcement Learning.
Himanshu SahniSaurabh KumarFarhan TejaniYannick SchroeckerCharles Lee Isbell Jr.Published in: CoRR (2017)
Keyphrases
- state space
- reinforcement learning
- transfer learning
- reinforcement learning algorithms
- optimal policy
- markov decision processes
- heuristic search
- action space
- function approximation
- state abstraction
- state variables
- markov chain
- dynamic programming
- decomposition method
- knowledge transfer
- partially observable
- continuous state spaces
- dynamical systems
- learning tasks
- state transition
- markov decision process
- learning agent
- goal state
- creation process
- temporal difference learning
- reward function
- supervised learning
- search space
- control problems
- state and action spaces
- markov decision problems
- decomposition algorithm
- machine learning
- deep learning
- temporal difference
- action selection
- infinite horizon
- finite state
- cross domain
- optimal control
- planning problems
- partial order
- learning process
- multi agent