Summary of Ho-fmn: Hyperparameter Optimization For Fast Minimum-norm Attacks, by Raffaele Mura et al.
HO-FMN: Hyperparameter Optimization for Fast Minimum-Norm Attacks
by Raffaele Mura, Giuseppe Floris, Luca Scionis, Giorgio Piras, Maura Pintor, Ambra Demontis, Giorgio Giacinto, Battista Biggio, Fabio Roli
First submitted to arxiv on: 11 Jul 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 The paper proposes a parametric variation of the fast minimum-norm attack algorithm to evaluate the robustness of machine-learning models. The authors address limitations in existing attacks by dynamically adjusting loss, optimizer, step-size scheduler, and hyperparameters. This allows for reporting adversarial robustness as a function of perturbation budget, providing a more complete evaluation than fixed-budget attacks. The approach is efficient and applicable to 12 robust models, achieving smaller adversarial perturbations without additional tuning. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper improves the way we test how well machine-learning models can handle being fooled by fake data. Right now, many of these tests use the same settings every time, which might not give a complete picture of how good or bad the model is. The authors came up with a new way to do these tests that lets them adjust certain settings on the fly. This makes their results more accurate and comprehensive. They tested this new approach on 12 different models and found it was better at finding small fake changes without needing extra fine-tuning. |
Keywords
» Artificial intelligence » Fine tuning » Machine learning
