Learning what to memorize: Using intrinsic motivation to form useful memory in partially observable reinforcement learning.
Alper DemirPublished in: Appl. Intell. (2023)
Keyphrases
- reinforcement learning
- partially observable
- intrinsic motivation
- learning process
- hidden state
- action models
- markov decision processes
- learning algorithm
- partially observable environments
- state space
- partially observable domains
- decision problems
- dynamical systems
- partial observations
- machine learning
- partial observability
- online learning
- function approximation
- temporal difference
- belief state
- markov decision problems
- orders of magnitude
- np hard