Data-Efficient Double-Win Lottery Tickets from Robust Pre-training

TL;DR

This paper introduces the concept of Double-Win Lottery Tickets, which are sparse subnetworks from pre-trained models that maintain both standard and adversarial robustness across diverse downstream tasks, especially benefiting from robust pre-training.

Contribution

It formulates the Double-Win Lottery Tickets concept, demonstrating their effectiveness and data efficiency, particularly when derived from robust pre-training methods.

Findings

Robust pre-training yields sparser and more effective double-win lottery tickets.

Double-win subnetworks perform well under both standard and adversarial training regimes.

These subnetworks are more data-efficient in transfer learning, especially with limited data.

Abstract

Pre-training serves as a broadly adopted starting point for transfer learning on various downstream tasks. Recent investigations of lottery tickets hypothesis (LTH) demonstrate such enormous pre-trained models can be replaced by extremely sparse subnetworks (a.k.a. matching subnetworks) without sacrificing transferability. However, practical security-crucial applications usually pose more challenging requirements beyond standard transfer, which also demand these subnetworks to overcome adversarial vulnerability. In this paper, we formulate a more rigorous concept, Double-Win Lottery Tickets, in which a located subnetwork from a pre-trained model can be independently transferred on diverse downstream tasks, to reach BOTH the same standard and robust generalization, under BOTH standard and adversarial training regimes, as the full pre-trained model can do. We comprehensively examine various pre-training mechanisms and find that robust pre-training tends to craft sparser double-win lottery tickets with superior performance over the standard counterparts. For example, on downstream CIFAR-10/100 datasets, we identify double-win matching subnetworks with the standard, fast adversarial, and adversarial pre-training from ImageNet, at 89.26%/73.79%, 89.26%/79.03%, and 91.41%/83.22% sparsity, respectively. Furthermore, we observe the obtained double-win lottery tickets can be more data-efficient to transfer, under practical data-limited (e.g., 1% and 10%) downstream schemes. Our results show that the benefits from robust pre-training are amplified by the lottery ticket scheme, as well as the data-limited transfer setting. Codes are available at https://github.com/VITA-Group/Double-Win-LTH.