Debiased Contrastive Learning
This addresses a key bottleneck in self-supervised representation learning for various AI domains, offering a practical improvement over existing methods.
The paper tackles the problem of sampling negative examples in contrastive learning that may have the same label, which harms performance, and proposes a debiased objective that corrects for this without needing true labels, achieving consistent state-of-the-art improvements across vision, language, and reinforcement learning benchmarks.
A prominent technique for self-supervised representation learning has been to contrast semantically similar and dissimilar pairs of samples. Without access to labels, dissimilar (negative) points are typically taken to be randomly sampled datapoints, implicitly accepting that these points may, in reality, actually have the same label. Perhaps unsurprisingly, we observe that sampling negative examples from truly different labels improves performance, in a synthetic setting where labels are available. Motivated by this observation, we develop a debiased contrastive objective that corrects for the sampling of same-label datapoints, even without knowledge of the true labels. Empirically, the proposed objective consistently outperforms the state-of-the-art for representation learning in vision, language, and reinforcement learning benchmarks. Theoretically, we establish generalization bounds for the downstream classification task.