Research map / Adversarial robustness and security / Certified and empirical robustness
Transferable Adversarial Robustness for Categorical Data via Universal Robust Embeddings
Klim Kireev, Maksym Andriushchenko, Carmela Troncoso, Nicolas Flammarion
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| apply universal robust embeddings | Random Forest · IEEECIS | robust accuracy (%): 48.0 -> 81.0 | Improved |
| apply universal robust embeddings | LGBM · IEEECIS | robust accuracy (%): 53.9 -> 78.5 | Improved |
| apply universal robust embeddings | Random Forest · BAF | robust accuracy (%): 42.8 -> 65.8 | Improved |
| apply universal robust embeddings | Random Forest · Credit | robust accuracy (%): 55.2 -> 66.7 | Improved |
| apply universal robust embeddings | LGBM · IEEECIS | clean accuracy (%): 81.2 -> 79.3 | Hurt |
| apply universal robust embeddings | LGBM · Credit | robust accuracy (%): 69.9 -> 72.5 | Improved |
| use robust embeddings instead of l1-trained boosted stumps (GBS-W) | Gradient Boosted Stumps · IEEECIS | robust accuracy (%): 11.1 -> 66.3 | Improved |
| merge close embeddings (tau=0.15) after bilevel training (ablation) | robust embeddings, GBS · BAF | robust accuracy (%): 48.3 -> 70.9 | Improved |
| merge close embeddings in random embeddings (ablation) | random embeddings, GBS · BAF | robust accuracy (%): 49.1 -> 44.7 | No clear effect |
| bilevel optimization instead of normal adversarial training for embeddings (tau=0.15) (ablation) | robust embeddings, GBS · BAF | robust accuracy (%): 53.3 -> 70.9 | Improved |
| use first-layer instead of last-layer embeddings (ablation) | LGBM · IEEECIS | clean accuracy (%): 78.9 -> 81.0 | Improved |
| use FT-Transformer embeddings instead of TabNet embeddings (ablation) | Gradient Boosted Stumps · BAF | robust accuracy (%): 67.7 -> 71.5 | Improved |
Closest papers by meaning
- TabularBench: Benchmarking Adversarial Robustness for Tabular Deep Learning in Real-world Use-cases (NeurIPS 2024 · similarity 0.90)
- Constrained Adaptive Attack: Effective Adversarial Attack Against Deep Neural Networks for Tabular Data (NeurIPS 2024 · similarity 0.88)
- Towards Understanding the Robustness Against Evasion Attack on Categorical Data (ICLR 2022 · similarity 0.88)
- Adversarial Robustness of Supervised Sparse Coding (NeurIPS 2020 · similarity 0.87)
- Attack-free Evaluating and Enhancing Adversarial Robustness on Categorical Data (ICML 2024 · similarity 0.87)
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
