Totally model-free actor-critic recurrent neural-network reinforcement learning in non-Markovian domains.
Eiji MizutaniStuart E. DreyfusPublished in: Ann. Oper. Res. (2017)
Keyphrases
- model free
- reinforcement learning
- recurrent neural networks
- actor critic
- reinforcement learning algorithms
- temporal difference
- policy iteration
- function approximation
- transfer learning
- rl algorithms
- neural network
- feed forward
- complex valued
- policy gradient
- approximate dynamic programming
- average reward
- state space
- temporal difference learning
- optimal control
- optimal policy
- dynamic programming
- learning process
- artificial neural networks
- learning algorithm
- neuro fuzzy
- markov decision processes
- reinforcement learning methods
- multi agent
- fuzzy logic
- partially observable
- markov decision process
- function approximators