DisCoRL: Continual Reinforcement Learning via Policy Distillation.
René TraoréHugo Caselles-DupréTimothée LesortTe SunGuanghang CaiNatalia Díaz RodríguezDavid FilliatPublished in: CoRR (2019)
Keyphrases
- reinforcement learning
- optimal policy
- policy search
- markov decision process
- action selection
- partially observable environments
- reinforcement learning algorithms
- function approximators
- policy gradient
- markov decision processes
- reinforcement learning problems
- control policies
- control policy
- approximate dynamic programming
- action space
- reward function
- partially observable
- policy iteration
- markov decision problems
- state space
- rl algorithms
- state and action spaces
- actor critic
- function approximation
- state action
- partially observable markov decision processes
- temporal difference
- optimal control
- approximate policy iteration
- learning algorithm
- partially observable domains
- agent learns
- average reward
- decision problems
- dynamic programming
- policy evaluation
- policy gradient methods
- continuous state spaces
- continuous state
- machine learning
- partially observable markov decision process
- policy making
- inverse reinforcement learning
- state dependent
- control problems
- learning problems
- search space
- neural network