TACO: Temporal Latent Action-Driven Contrastive Loss for Visual Reinforcement Learning.
Ruijie ZhengXiyao WangYanchao SunShuang MaJieyu ZhaoHuazhe XuHal Daumé IIIFurong HuangPublished in: CoRR (2023)
Keyphrases
- reinforcement learning
- action selection
- action space
- visual information
- temporal information
- data driven
- partially observable domains
- visual features
- spatio temporal
- spatial and temporal
- temporal data
- function approximation
- temporal abstractions
- sensory inputs
- visual perception
- temporal reasoning
- temporal databases
- human actions
- latent variables
- markov decision processes
- low level
- reinforcement learning algorithms
- temporal relations
- dynamic programming
- reward shaping
- machine learning
- temporal constraints
- optimal policy
- temporal difference
- action models
- high level
- transition model
- learning algorithm