Pruning Convolutional Neural Networks with Self-Supervision
This addresses the computational inefficiency of large unsupervised convnets for researchers and practitioners, but it is incremental as it applies existing pruning methods to self-supervised learning.
The paper tackled the problem of pruning convolutional neural networks trained with self-supervision to reduce computational costs while preserving performance on downstream tasks, finding that standard pruning methods work similarly for self-supervised and supervised learning and maintain transfer performance.
Convolutional neural networks trained without supervision come close to matching performance with supervised pre-training, but sometimes at the cost of an even higher number of parameters. Extracting subnetworks from these large unsupervised convnets with preserved performance is of particular interest to make them less computationally intensive. Typical pruning methods operate during training on a task while trying to maintain the performance of the pruned network on the same task. However, in self-supervised feature learning, the training objective is agnostic on the representation transferability to downstream tasks. Thus, preserving performance for this objective does not ensure that the pruned subnetwork remains effective for solving downstream tasks. In this work, we investigate the use of standard pruning methods, developed primarily for supervised learning, for networks trained without labels (i.e. on self-supervised tasks). We show that pruned masks obtained with or without labels reach comparable performance when re-trained on labels, suggesting that pruning operates similarly for self-supervised and supervised learning. Interestingly, we also find that pruning preserves the transfer performance of self-supervised subnetwork representations.