Actor-Critic Policy Optimization in Partially Observable Multiagent Environments.
Sriram SrinivasanMarc LanctotVinícius Flores ZambaldiJulien PérolatKarl TuylsRémi MunosMichael BowlingPublished in: NeurIPS (2018)
Keyphrases
- partially observable
- actor critic
- reinforcement learning
- markov decision processes
- policy iteration
- state space
- markov decision problems
- policy gradient
- partially observable markov decision processes
- infinite horizon
- dynamical systems
- decision problems
- reinforcement learning algorithms
- optimal control
- optimal policy
- average reward
- temporal difference
- reward function
- belief state
- function approximation
- single agent
- approximate dynamic programming
- learning agent
- multi agent
- neuro fuzzy
- optimization problems
- markov decision process
- dynamic programming
- learning algorithm
- model free
- average cost
- semi supervised
- search space
- action selection
- control policy
- genetic algorithm
- decision making
- finite state
- heuristic search