Non-asymptotic Convergence of Adam-type Reinforcement Learning Algorithms under Markovian Sampling.
Huaqing XiongTengyu XuYingbin LiangWei ZhangPublished in: AAAI (2021)
Keyphrases
- reinforcement learning algorithms
- asymptotic convergence
- reinforcement learning
- state space
- model free
- markov decision processes
- reinforcement learning problems
- reinforcement learning methods
- temporal difference
- function approximation
- learning algorithm
- dynamic environments
- monte carlo
- particle swarm algorithm
- natural gradient
- neural network
- stochastic games
- markov chain
- artificial neural networks