Contrastive Learning Is Spectral Clustering On Similarity Graph
This provides a foundational theoretical insight for researchers in self-supervised learning, though it is incremental in method development.
The paper tackles the limited theoretical understanding of contrastive learning by proving it is equivalent to spectral clustering on a similarity graph, and introduces a Kernel-InfoNCE loss that outperforms standard methods on vision datasets with concrete improvements.
Contrastive learning is a powerful self-supervised learning method, but we have a limited theoretical understanding of how it works and why it works. In this paper, we prove that contrastive learning with the standard InfoNCE loss is equivalent to spectral clustering on the similarity graph. Using this equivalence as the building block, we extend our analysis to the CLIP model and rigorously characterize how similar multi-modal objects are embedded together. Motivated by our theoretical insights, we introduce the Kernel-InfoNCE loss, incorporating mixtures of kernel functions that outperform the standard Gaussian kernel on several vision datasets. The code is available at https://github.com/yifanzhang-pro/Kernel-InfoNCE.