A Discussion of 'Adversarial Examples Are Not Bugs, They Are Features': Two Examples of Useful, Non-Robust Features

🔬 Curated from Distill.pub Read original →

DeepTrendLab's Take on A Discussion of 'Adversarial Examples Are Not Bugs, They...

Neural networks rely on non-robust features that correlate with labels but fail under adversarial attack. Research explores how these fragile patterns emerge even in simple linear models.

This article was originally published on Distill.pub. Read the full piece at the source.

Read full article on Distill.pub →

DeepTrendLab curates AI news from 50+ sources. All original content and rights belong to Distill.pub. DeepTrendLab's analysis is independently written and does not represent the views of the original publisher.