Offline reinforcement learning with representations for actions.
Xingzhou LouQiyue YinJunge ZhangChao YuZhaofeng HeNengjie ChengKaiqi HuangPublished in: Inf. Sci. (2022)
Keyphrases
- reinforcement learning
- perceptual aliasing
- action selection
- partially observable
- action space
- reward function
- function approximation
- state and action spaces
- partial observability
- multiagent reinforcement learning
- partially observable domains
- state space
- real time
- learned knowledge
- learning agent
- function approximators
- temporal difference
- agent receives
- machine learning
- dynamic programming
- model free
- decision theoretic
- learning classifier systems
- state action
- action models
- learning process
- robotic control
- optimal policy
- agent learns
- conceptual representation
- multiagent systems
- initially unknown
- markov decision processes
- learning problems
- human actions
- situation calculus
- policy iteration
- plan recognition
- symbolic representation
- goal directed