MDP Homomorphic Networks: Group Symmetries in Reinforcement Learning.
Elise van der PolDaniel E. WorrallHerke van HoofFrans A. OliehoekMax WellingPublished in: NeurIPS (2020)
Keyphrases
- reinforcement learning
- markov decision processes
- optimal policy
- markov decision process
- state space
- reinforcement learning algorithms
- state and action spaces
- function approximation
- partially observable
- action sets
- action space
- multi agent
- dynamic programming
- privacy preserving
- markov decision problems
- complex networks
- supervised learning
- temporal difference
- policy iteration
- real time dynamic programming
- heterogeneous networks
- reward function
- finite state
- utility function
- social networks
- learning algorithm
- partially observable markov decision processes
- markov chain
- state abstraction
- policy search
- bayesian reinforcement learning
- neural network