Constrained Reinforcement Learning via Policy Splitting.
Haoxian ChenHenry LamFengpei LiAmirhossein MeisamiPublished in: ACML (2020)
Keyphrases
- reinforcement learning
- optimal policy
- policy search
- action selection
- markov decision process
- partially observable environments
- reinforcement learning algorithms
- action space
- function approximation
- function approximators
- actor critic
- markov decision processes
- control policy
- state space
- approximate dynamic programming
- state and action spaces
- policy gradient
- reinforcement learning problems
- reward function
- control policies
- decision problems
- policy iteration
- dynamic programming
- state action
- learning algorithm
- policy evaluation
- markov decision problems
- rl algorithms
- model free
- average reward
- long run
- finite state
- state dependent
- agent learns
- partially observable
- policy gradient methods
- optimal control
- temporal difference
- neural network
- exploration exploitation tradeoff
- eligibility traces
- robotic control
- partially observable domains
- inverse reinforcement learning
- continuous state spaces
- policy making
- continuous state
- reinforcement learning methods
- partially observable markov decision processes
- infinite horizon
- objective function
- machine learning