Summary of Adversarial Robustness Of Vaes Across Intersectional Subgroups, by Chethan Krishnamurthy Ramanaik et al.
Adversarial Robustness of VAEs across Intersectional Subgroups
by Chethan Krishnamurthy Ramanaik, Arjun Roy, Eirini Ntoutsi
First submitted to arxiv on: 4 Jul 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI)
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 investigates the robustness of Variational Autoencoders (VAEs) against non-targeted adversarial attacks. Despite VAEs’ stronger resistance compared to deterministic AEs, they are still vulnerable to such perturbations. The study optimizes minimal sample-specific perturbations to cause maximal damage across diverse demographic subgroups and examines factors contributing to robustness disparities like data scarcity and representation entanglement. Results show that robustness disparities exist but are not always correlated with subgroup size, highlighting the vulnerability of subgroups like older women to misclassification due to adversarial perturbations. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This study looks at how well Variational Autoencoders (VAEs) can handle bad data. Even though VAEs are better than other types of autoencoders, they’re still not perfect. The researchers wanted to see if some groups of people were more likely to be fooled by bad data than others. They found that yes, some groups are more vulnerable, like older women. |
