Investigating the Edge of Stability Phenomenon in Reinforcement Learning.
Rares IordanMarc Peter DeisenrothMihaela RoscaPublished in: CoRR (2023)
Keyphrases
- reinforcement learning
- stability analysis
- function approximation
- state space
- edge detection
- reinforcement learning algorithms
- edge information
- machine learning
- markov decision processes
- neural network
- action selection
- edge detector
- optimal policy
- dynamic programming
- multi agent
- optimal control
- weighted graph
- learning classifier systems
- model free
- learning algorithm
- markov decision process
- learning environment
- image segmentation
- genetic algorithm
- partially observable
- database
- temporal difference learning
- transition model