A Meta-MDP Approach to Exploration for Lifelong Reinforcement Learning.
Francisco M. GarciaPhilip S. ThomasPublished in: NeurIPS (2019)
Keyphrases
- reinforcement learning
- exploration strategy
- markov decision processes
- model based reinforcement learning
- markov decision process
- exploration exploitation
- optimal policy
- state space
- action selection
- active exploration
- action sets
- function approximation
- partially observable
- reward function
- learning algorithm
- state and action spaces
- model free
- balancing exploration and exploitation
- machine learning
- temporal difference
- dynamic programming
- markov decision problems
- exploration exploitation tradeoff
- autonomous learning
- policy iteration
- finite state
- reinforcement learning algorithms
- optimal control
- technology enhanced learning
- bayesian reinforcement learning
- learning activities
- unknown environments
- action space
- supervised learning
- st century
- multi agent
- planning under uncertainty
- meta level
- partial observability
- lifelong learning
- transition model
- search algorithm
- decision problems