Research map / Adversarial robustness and security / Certified and empirical robustness
Enhancing Adversarial Robustness for Deep Metric Learning
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use HM with LGA and add the ICS loss (ablation) | ResNet-18 deep metric learning · CUB | ERS: 32.4 -> 33.5 | Helped |
| Use HM with LGA and add the ICS loss | ResNet-18 deep metric learning · Cars-196 | ERS: 42.4 -> 47.9 | Helped |
| Use HM with LGA and add the ICS loss | ResNet-18 deep metric learning · SOP | ERS: 61.7 -> 61.2 | No clear effect |
| Use LGA destination hardness instead of sampled Semihard triplet hardness (ablation) | HM[Softhard] on CUB · CUB | ERS: 29.6 -> 32.4 | Helped |
| Replace LGA with fixed destination hardness -γ/2 (ablation) | HM[Softhard] on CUB · CUB | ERS: 32.4 -> 33.8 | Helped |
| Replace LGA with fixed destination hardness -γ/2 (ablation) | HM[Softhard] on CUB · CUB | R@1: 38.0 -> 36.8 | Hurt |
| Set destination hardness to 0 (ablation) | HM[Softhard] on CUB · CUB | R@1: 38.0 -> 0.8 | Hurt |
| Slightly boost destination hardness with g_B (ablation) | HM[Softhard] on CUB · CUB | ERS: 29.6 -> 33.8 | Helped |
| Slightly boost destination hardness with g_B (ablation) | HM[Softhard] on CUB · CUB | R@1: 38.4 -> 36.5 | Hurt |
| Use LGA with clipping parameter u=2.2 rather than the default (ablation) | HM[Softhard] on CUB · CUB | R@1: 38.0 -> 34.8 | Hurt |
| Use LGA with clipping parameter u=2.2 rather than the default (ablation) | HM[Softhard] on CUB · CUB | ERS: 32.4 -> 36.1 | Helped |
| Increase ICS weight to λ=1.0 (ablation) | HM[Softhard, LGA] + ICS on CUB · CUB | R@1: 37.2 -> 36.0 | Hurt |
Closest papers by meaning
- Metric Learning for Adversarial Robustness (NeurIPS 2019 · similarity 0.91)
- On the Duality Between Sharpness-Aware Minimization and Adversarial Training (ICML 2024 · similarity 0.87)
- Provably Cost-Sensitive Adversarial Defense via Randomized Smoothing (ICML 2025 · similarity 0.86)
- Comparing the Robustness of Modern No-Reference Image- and Video-Quality Metrics to Adversarial Attacks (AAAI 2024 · similarity 0.86)
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.86)
- Collapse-Aware Triplet Decoupling for Adversarially Robust Image Retrieval (ICML 2024 · similarity 0.85)
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.