Concave Utility Reinforcement Learning: the Mean-field Game viewpoint.
Matthieu GeistJulien PérolatMathieu LaurièreRomuald ElieSarah PerrinOlivier BachemRémi MunosOlivier PietquinPublished in: CoRR (2021)
Keyphrases
- viewpoint
- reinforcement learning
- game theory
- computer games
- video games
- game playing
- markov random field
- machine learning
- objective function
- temporal difference
- piecewise linear
- function approximation
- learning algorithm
- reinforcement learning algorithms
- multiple views
- utility function
- closed form
- model free
- virtual world
- illumination conditions
- game theoretic
- game play
- transferable utility
- temporal difference learning
- state space
- em algorithm
- serious games
- educational games
- game design
- bayesian inference
- nash equilibrium
- belief networks
- action selection
- online game
- reinforcement learning methods
- markov networks
- d objects
- optimal policy
- dynamic programming
- stochastic games
- fixed point
- markov decision processes