Provable Guarantees for Self-Supervised Deep Learning with Spectral Contrastive Loss

Abstract

Recent works in self-supervised learning have advanced the state-of-the-art by relying on the contrastive learning paradigm, which learns representations by pushing positive pairs, or similar examples from the same class, closer together while keeping negative pairs far apart. Despite the empirical successes, theoretical foundations are limited -- prior analyses assume conditional independence of the positive pairs given the same class label, but recent empirical applications use heavily correlated positive pairs (i.e., data augmentations of the same image). Our work analyzes contrastive learning without assuming conditional independence of positive pairs using a novel concept of the augmentation graph on data. Edges in this graph connect augmentations of the same data, and ground-truth classes naturally form connected sub-graphs. We propose a loss that performs spectral decomposition on the population augmentation graph and can be succinctly written as a contrastive learning objective on neural net representations. Minimizing this objective leads to features with provable accuracy guarantees under linear probe evaluation. By standard generalization bounds, these accuracy guarantees also hold when minimizing the training contrastive loss. Empirically, the features learned by our objective can match or outperform several strong baselines on benchmark vision datasets. In all, this work provides the first provable analysis for contrastive learning where guarantees for linear probe evaluation can apply to realistic empirical settings.

Accepted as an oral to NeurIPS 2021

References in corpus (9)

• Bootstrap your own latent: A new approach to self-supervised Learning
• Barlow Twins: Self-Supervised Learning via Redundancy Reduction
• Exploring Simple Siamese Representation Learning
• Theoretical Analysis of Self-Training with Deep Networks on Unlabeled Data
• The geometry of kernelized spectral clustering
• Understanding Self-supervised Learning with Dual Deep Networks
• Contrastive Learning Inverts the Data Generating Process
• Contrastive estimation reveals topic posterior information to linear models
• A Theory of Label Propagation for Subpopulation Shift

Cited by in corpus (5)

• On the Opportunities and Risks of Foundation Models
• Self-supervised Learning is More Robust to Dataset Imbalance
• Unifying Graph Convolution and Contrastive Learning in Collaborative Filtering
• A Co-training Approach for Noisy Time Series Learning
• The Power of Contrast for Feature Learning: A Theoretical Analysis