Verifiable Reinforcement Learning via Policy Extraction.
Osbert BastaniYewen PuArmando Solar-LezamaPublished in: NeurIPS (2018)
Keyphrases
- reinforcement learning
- optimal policy
- policy search
- markov decision process
- action selection
- actor critic
- sufficient conditions
- policy iteration
- partially observable environments
- information extraction
- function approximators
- state space
- reinforcement learning problems
- partially observable
- markov decision problems
- function approximation
- automatic extraction
- reinforcement learning algorithms
- control policy
- approximate dynamic programming
- temporal difference
- reward function
- markov decision processes
- control policies
- state action
- machine learning
- inverse reinforcement learning
- state and action spaces
- temporal difference learning
- partially observable domains
- model free
- dynamic programming
- secret sharing
- action space
- continuous state
- learning process
- policy gradient
- agent learns
- infinite horizon
- long run
- decision problems
- decision making
- approximate policy iteration
- multi agent
- supervised learning
- continuous state spaces
- policy evaluation
- policy making
- control problems
- rl algorithms
- state dependent