Studying the impact of magnitude pruning on contrastive learning methods
This work addresses the problem of optimizing pruning techniques for contrastive learning methods, which is incremental as it builds on existing pruning and representation quality metrics.
The study investigated how magnitude pruning affects contrastive learning in deep neural networks, finding that at high sparsity levels, contrastive learning leads to more misclassified examples compared to cross-entropy loss, with early pruning causing the greatest negative impact on representation quality.
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.