Policy Consolidation for Continual Reinforcement Learning.
Christos KaplanisMurray ShanahanClaudia ClopathPublished in: CoRR (2019)
Keyphrases
- reinforcement learning
- optimal policy
- policy search
- markov decision process
- action selection
- partially observable environments
- function approximators
- markov decision processes
- control policy
- function approximation
- policy iteration
- partially observable
- actor critic
- markov decision problems
- control policies
- state space
- policy gradient
- continuous state
- approximate dynamic programming
- reinforcement learning algorithms
- reinforcement learning problems
- action space
- policy evaluation
- state action
- state and action spaces
- state dependent
- decision problems
- reward function
- partially observable markov decision processes
- rl algorithms
- model free
- multi agent
- temporal difference learning
- temporal difference
- infinite horizon
- finite state
- learning algorithm
- allocation policy
- agent learns
- learning problems
- optimal control
- learning capabilities
- average reward
- long run
- inverse reinforcement learning
- dynamic pricing
- dynamic programming
- average cost
- eligibility traces
- robotic control
- policy makers
- approximate policy iteration
- partially observable domains
- natural actor critic
- policy gradient methods