Studying the impact of magnitude pruning on contrastive learning methods

TL;DR

This paper investigates how different pruning techniques affect the quality of representations learned by contrastive learning models, revealing that early pruning at high sparsity levels degrades performance more than traditional training methods.

Contribution

It provides a detailed analysis of the effects of pruning schedules on contrastive learning, highlighting the importance of pruning timing for maintaining representation quality.

Findings

High sparsity in contrastive learning increases misclassification.

Early pruning during training significantly worsens representation quality.

Metrics like PIEs, Q-Score, and PD-Score quantify pruning impact.

Abstract

We study the impact of different pruning techniques on the representation learned by deep neural networks trained with contrastive loss functions. Our work finds that at high sparsity levels, contrastive learning results in a higher number of misclassified examples relative to models trained with traditional cross-entropy loss. To understand this pronounced difference, we use metrics such as the number of PIEs (Hooker et al., 2019), Q-Score (Kalibhat et al., 2022), and PD-Score (Baldock et al., 2021) to measure the impact of pruning on the learned representation quality. Our analysis suggests the schedule of the pruning method implementation matters. We find that the negative impact of sparsity on the quality of the learned representation is the highest when pruning is introduced early on in the training phase.