Deep Episodic Value Iteration for Model-based Meta-Reinforcement Learning.
Steven Stenberg HansenPublished in: CoRR (2017)
Keyphrases
- reinforcement learning
- markov decision processes
- state space
- model free
- policy iteration
- optimal policy
- markov decision process
- dynamic programming
- reinforcement learning algorithms
- partially observable markov decision processes
- meta level
- average reward
- heuristic search
- temporal difference
- function approximation
- machine learning
- markov decision chains
- neural network
- action space
- meta reasoning
- factored mdps
- belief space
- reinforcement learning methods
- temporal difference learning
- learning algorithm
- data driven
- finite state
- deep learning
- infinite horizon
- multi agent
- search algorithm
- approximate dynamic programming
- multi agent reinforcement learning
- reward function
- dynamical systems