STACoRe: Spatio-temporal and action-based contrastive representations for reinforcement learning in Atari.
Young Jae LeeJaehoon KimMingu KwakYoung Joon ParkSeoung Bum KimPublished in: Neural Networks (2023)
Keyphrases
- spatio temporal
- reinforcement learning
- linear value function approximation
- human actions
- action selection
- reinforcement learning problems
- action space
- function approximation
- action recognition
- agent learns
- reinforcement learning algorithms
- learning algorithm
- optimal policy
- state space
- multi agent
- image sequences
- dynamic programming
- temporal difference learning
- machine learning
- model free
- learning process
- moving objects
- function approximators
- reinforcement learning methods
- optimal control
- temporal difference
- learning capabilities
- control problems