Summary of Adversarial Fine-tuning Of Compressed Neural Networks For Joint Improvement Of Robustness and Efficiency, by Hallgrimur Thorsteinsson et al.
Adversarial Fine-tuning of Compressed Neural Networks for Joint Improvement of Robustness and Efficiency
by Hallgrimur Thorsteinsson, Valdemar J Henriksen, Tong Chen, Raghavendra Selvan
First submitted to arxiv on: 14 Mar 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary Deep learning (DL) models are increasingly being integrated into our everyday lives, making it crucial to ensure their safety by making them robust against adversarial attacks. Adversarial training has been presented as a mitigation strategy, but this approach comes with additional computational costs required to design adversarial attacks during training. The two objectives – adversarial robustness and computational efficiency – then appear to be in conflict with each other. To address this issue, we explored the effects of two different model compression methods (structured weight pruning and quantization) on adversarial robustness. We specifically explored the effects of fine-tuning on compressed models, presenting a trade-off between standard fine-tuning and adversarial fine-tuning. Our results show that compression does not inherently lead to loss in model robustness, and adversarial fine-tuning of a compressed model can yield large improvement to the robustness performance of models. We presented experiments on two benchmark datasets showing that adversarial fine-tuning of compressed models can achieve robustness performance comparable to adversarially trained models, while also improving computational efficiency. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Deep learning models are super important in our daily lives, but they’re really vulnerable to bad guys trying to mess with them. To fix this, we tried two new ways to make the models stronger (structured weight pruning and quantization). We wanted to see if these methods would work even better if we fine-tuned them (made small changes) for extra protection. Our results show that making the models smaller doesn’t mean they’re weaker – it can actually help keep them safe! We tested this on two big datasets and found out that making the compressed models stronger can be just as good as training them with special bad-guy attacks, but it’s way more efficient. |
Keywords
* Artificial intelligence * Deep learning * Fine tuning * Model compression * Pruning * Quantization
