Why do We Need Large Batchsizes in Contrastive Learning? A Gradient-Bias Perspective.
Changyou ChenJianyi ZhangYi XuLiqun ChenJiali DuanYiran ChenSon TranBelinda ZengTrishul ChilimbiPublished in: NeurIPS (2022)
Keyphrases
- learning algorithm
- learning scheme
- learning process
- machine learning
- reinforcement learning
- background knowledge
- average reward reinforcement learning
- multiple perspectives
- inductive inference
- inductive learning
- learning tasks
- online learning
- learning systems
- prior knowledge
- mobile learning
- trade off
- feature space
- learning analytics
- decision trees
- knowledge base
- genetic algorithm
- database