Concave Utility Reinforcement Learning: The Mean-field Game Viewpoint.
Matthieu GeistJulien PérolatMathieu LaurièreRomuald ElieSarah PerrinOlivier BachemRémi MunosOlivier PietquinPublished in: AAMAS (2022)
Keyphrases
- viewpoint
- reinforcement learning
- function approximation
- computer games
- video games
- objective function
- utility function
- game theory
- markov decision processes
- temporal difference
- game playing
- markov networks
- markov random field
- multi agent
- reinforcement learning algorithms
- game play
- educational games
- game design
- game theoretic
- imperfect information
- stochastic games
- nash equilibrium
- multiple views
- learning algorithm
- learning process
- virtual world
- em algorithm
- loopy belief propagation
- function approximators
- online game
- state space
- optimal policy
- piecewise linear
- closed form
- belief networks
- multiagent reinforcement learning
- dynamic programming
- transferable utility
- graphical models
- linear complexity
- variational methods
- action selection
- model free
- bayesian inference