Self-organization of action hierarchy and compositionality by reinforcement learning with recurrent neural networks.
Dongqi HanKenji DoyaJun TaniPublished in: Neural Networks (2020)
Keyphrases
- recurrent neural networks
- reinforcement learning
- action selection
- neural network
- partially observable domains
- multi agent
- complex valued
- reward shaping
- action space
- feed forward
- function approximation
- state action
- reservoir computing
- neural model
- state space
- recurrent networks
- artificial neural networks
- echo state networks
- cascade correlation
- feedforward neural networks
- hebbian learning
- reinforcement learning algorithms
- machine learning
- learning algorithm
- data mining
- agent receives
- model free
- transition model
- agent learns
- long term
- dynamic environments
- biologically inspired
- transfer learning