Biologically Inspired Reinforcement Learning: Reward-Based Decomposition for Multi-goal Environments.
Weidong ZhouRichard CogginsPublished in: BioADIT (2004)
Keyphrases
- biologically inspired
- reinforcement learning
- receptive fields
- biologically plausible
- function approximation
- potential field
- biological systems
- scene classification
- reward function
- reinforcement learning algorithms
- dynamic environments
- learning rules
- agent learns
- state space
- reinforcement learning methods
- sensory inputs
- recurrent networks
- learning algorithm
- humanoid robot
- eligibility traces
- neural network
- model free
- optimal policy
- spiking neural networks
- temporal difference
- dynamic programming
- center surround
- policy gradient
- reward signal
- initially unknown
- visual attention model
- mobile robot
- average reward
- control policy
- learning agent
- markov decision process