Contrastive quant: quantization makes stronger contrastive learning
Contrastive quant: quantization makes stronger contrastive learning
复制标题
DOI：
10.1145/3489517.3530419
复制
发表时间：
2022-07
期刊：
影响因子：
--
通讯作者：
Y. Fu;Qixuan Yu;Meng Li;Xuefeng Ouyang;Vikas Chandra;Yingyan Lin
中科院分区：
文献类型：
--
作者：
Y. Fu;Qixuan Yu;Meng Li;Xuefeng Ouyang;Vikas Chandra;Yingyan Lin
对比学习通过在不同的增强视图下强制特征一致性来学习视觉表征。在这项工作中，我们从一个新的角度探索对比学习。有趣的是，我们发现，量化，如果设计得当，可以提高对比学习的有效性。为此，我们提出了一种新的对比学习框架，称为对比量化，鼓励通过各种数据变换和不同的增强权重/激活通过各种量化水平下的不同增强输入的特征一致性。建立在两种最先进的对比学习方法Simplified和BYOL之上的大量实验表明，对比量化始终提高了学习到的视觉表征。
Contrastive learning learns visual representations by enforcing feature consistency under different augmented views. In this work, we explore contrastive learning from a new perspective. Interestingly, we find that quantization, when properly engineered, can enhance the effectiveness of contrastive learning. To this end, we propose a novel contrastive learning framework, dubbed Contrastive Quant, to encourage feature consistency under both differently augmented inputs via various data transformations and differently augmented weights/activations via various quantization levels. Extensive experiments, built on top of two state-of-the-art contrastive learning methods SimCLR and BYOL, show that Contrastive Quant consistently improves the learned visual representation.