Look where you look! Saliency-guided Q-networks for generalization in visual Reinforcement Learning.
David BertoinAdil ZouitineMehdi ZouitineEmmanuel RachelsonPublished in: NeurIPS (2022)
Keyphrases
- reinforcement learning
- visual saliency
- scan path
- visual information
- human visual
- markov decision processes
- neural network
- function approximation
- natural images
- network structure
- complex networks
- reinforcement learning algorithms
- heterogeneous networks
- visual perception
- optimal control
- state space
- machine learning
- information content
- visual attention
- action selection
- low level
- video sequences
- high level