Verifiable Reinforcement Learning via Policy Extraction.
Osbert BastaniYewen PuArmando Solar-LezamaPublished in: CoRR (2018)
Keyphrases
- reinforcement learning
- optimal policy
- policy search
- markov decision processes
- markov decision process
- partially observable environments
- reinforcement learning problems
- action selection
- function approximation
- reinforcement learning algorithms
- partially observable
- sufficient conditions
- policy iteration
- state space
- information extraction
- markov decision problems
- policy evaluation
- function approximators
- action space
- actor critic
- dynamic programming
- approximate dynamic programming
- learning algorithm
- asymptotically optimal
- temporal difference
- automatic extraction
- continuous state spaces
- reward function
- transfer learning
- model free
- average reward
- state action
- infinite horizon
- rl algorithms
- multi agent
- policy gradient
- state and action spaces
- policy making
- partially observable domains
- control policies
- temporal difference learning
- policy makers
- partially observable markov decision processes
- agent receives
- policy gradient methods
- model free reinforcement learning
- reinforcement learning methods
- agent learns
- learning process
- multi agent reinforcement learning
- continuous state
- learning problems
- least squares
- optimal control