Contrastive Learning with Hard Negative Samples
Contrastive Learning with Hard Negative Samples
复制标题
DOI：
--
复制
发表时间：
2020-10
期刊：
影响因子：
--
通讯作者：
Joshua Robinson;Ching-Yao Chuang;S. Sra;S. Jegelka
中科院分区：
文献类型：
--
作者：
Joshua Robinson;Ching-Yao Chuang;S. Sra;S. Jegelka
你如何为对比学习找出好的反面例子？我们认为，与度量学习一样，表征的对比学习受益于硬负样本（即难以与锚点区分的点）。使用硬否定的关键挑战是，对比方法必须保持无监督，这使得采用使用真实相似信息的现有负抽样策略变得不可行的。因此，我们开发了一种新的无监督采样方法，用于选择用户可以控制硬度的硬负样本。这种采样的一种极限情况会导致将每个类紧密聚类，并将不同的类尽可能地分开。所提出的方法提高了跨多种模式的下游性能，只需要几行额外的代码来实现，并且没有引入计算开销。
How can you sample good negative examples for contrastive learning? We argue that, as with metric learning, contrastive learning of representations benefits from hard negative samples (i.e., points that are difficult to distinguish from an anchor point). The key challenge toward using hard negatives is that contrastive methods must remain unsupervised, making it infeasible to adopt existing negative sampling strategies that use true similarity information. In response, we develop a new family of unsupervised sampling methods for selecting hard negative samples where the user can control the hardness. A limiting case of this sampling results in a representation that tightly clusters each class, and pushes different classes as far apart as possible. The proposed method improves downstream performance across multiple modalities, requires only few additional lines of code to implement, and introduces no computational overhead.