Assured Reinforcement Learning with Formally Verified Abstract Policies.
George MasonRadu CalinescuDaniel KudenkoAlec BanksPublished in: ICAART (2) (2017)
Keyphrases
- reinforcement learning
- optimal policy
- policy search
- markov decision process
- control policies
- hierarchical reinforcement learning
- state space
- decision problems
- markov decision processes
- function approximation
- control policy
- partially observable markov decision processes
- fitted q iteration
- total reward
- cooperative multi agent systems
- policy gradient methods
- model free
- machine learning
- reward function
- dynamic programming
- markov decision problems
- continuous state
- robotic control
- partially observable
- reinforcement learning agents
- decentralized control
- neural network
- genetic algorithm
- learning process
- revenue management
- temporal difference
- action space
- reinforcement learning algorithms