Noisy Symbolic Abstractions for Deep RL: A case study with Reward Machines.
Andrew C. LiZizhao ChenPashootan VaezipoorToryn Q. KlassenRodrigo Toro IcarteSheila A. McIlraithPublished in: CoRR (2022)
Keyphrases
- reinforcement learning
- high level
- noisy data
- state space
- temporal abstractions
- reinforcement learning algorithms
- reward function
- function approximation
- test bed
- symbolic description
- multi agent
- long run
- noisy environments
- eligibility traces
- policy gradient
- genetic algorithm
- learning machines
- model free
- state action
- noise free
- symbolic representation
- action selection
- real valued
- missing data
- supervised learning
- dynamic programming
- learning process
- case study