Optimal transport improves contrastive learning for visual representations
Positive Pair Geometry Matters: Optimal Transport for Contrastive Learning of Visual Representations
Computer Vision and Pattern RecognitionComputational Geometry
Summary
Self-supervised learning teaches computers to understand images by comparing different versions of the same picture. Usually, these versions are made by random changes that might accidentally change important details. The authors created a method that uses a math tool called optimal transport to blend the original and changed images in a smooth way. This helps keep the important parts consistent and makes the computer learn better. Tests show this method improves how well the computer understands and transfers image knowledge.
What this means in practice
- •For computer vision engineers: Enhance image feature models by integrating geometry-aware positive sample generation during training for improved accuracy and transferability.
- •For machine learning platform developers: Incorporate optimal transport-based interpolation techniques into contrastive learning pipelines without changing existing encoder architectures.
Authors
Akshit Nanda, Shahzad Ahmad, Ram Prasad Padhy
Abstract
Contrastive self-supervised learning has achieved strong performance by learning representations from multiple augmented views of the same image. However, most existing methods construct positive pairs using independently sampled stochastic augmentations, which may alter semantic content and ignore the intrinsic geometry of the data distribution. In this work, we propose OTCLR, an optimal transport-aware framework for contrastive learning representations that generates geometry-consistent positive samples. Instead of directly contrasting two randomly augmented views, we construct intermediate views between the original image and its augmented variants through entropic optimal-transport displacement interpolation. These transport-interpolated samples serve as positive views that better preserve image structure while explicitly modeling spatial distributional geometry. To further promote smooth representation learning, we evaluate auxiliary Sinkhorn regularization terms that encourage transport-interpolated views to remain consistent with their endpoint images. The proposed method can be incorporated into standard contrastive learning pipelines without modifying the encoder architecture. Experiments on multiple benchmark datasets show that our approach improves representation quality and transfer learning performance compared with conventional augmentation-based contrastive learning baselines.