Visualizing and Understanding Contrastive Learning

TL;DR

This paper develops and evaluates visual explanation methods for contrastive learning models, addressing the unique challenges of explaining paired inputs and assessing their effectiveness in understanding model behavior.

Contribution

It introduces new visual explanation techniques tailored for contrastive learning and adapts evaluation metrics to better assess explanations of paired data.

Findings

Proposed explanation methods effectively highlight important features in image pairs.

Evaluation metrics correlate explanations with downstream task performance.

Analysis reveals strengths and limitations of current explainability approaches for contrastive models.

Abstract

Contrastive learning has revolutionized the field of computer vision, learning rich representations from unlabeled data, which generalize well to diverse vision tasks. Consequently, it has become increasingly important to explain these approaches and understand their inner workings mechanisms. Given that contrastive models are trained with interdependent and interacting inputs and aim to learn invariance through data augmentation, the existing methods for explaining single-image systems (e.g., image classification models) are inadequate as they fail to account for these factors and typically assume independent inputs. Additionally, there is a lack of evaluation metrics designed to assess pairs of explanations, and no analytical studies have been conducted to investigate the effectiveness of different techniques used to explaining contrastive learning. In this work, we design visual explanation methods that contribute towards understanding similarity learning tasks from pairs of images. We further adapt existing metrics, used to evaluate visual explanations of image classification systems, to suit pairs of explanations and evaluate our proposed methods with these metrics. Finally, we present a thorough analysis of visual explainability methods for contrastive learning, establish their correlation with downstream tasks and demonstrate the potential of our approaches to investigate their merits and drawbacks.