Research map / Adversarial robustness and security / Certified and empirical robustness
On the Versatile Uses of Partial Distance Correlation in Deep Learning
Xingjian Zhen, Zihang Meng, Rudrasis Chakraborty, Vikas Singh
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 |
|---|---|---|---|
| add distance correlation loss against a reference network | ResNet-18 on CIFAR10 · PGD eps=0.1 transfer attack | accuracy (%): 27.41 -> 36.11 | Improved |
| add distance correlation loss against a reference network | ResNet-18 on CIFAR10 · PGD eps=0.05 transfer attack | accuracy (%): 49.42 -> 59.33 | Improved |
| add distance correlation loss against a reference network | MobileNet-v3-small on ImageNet · PGD eps=0.1 transfer attack | accuracy (%): 17.15 -> 28.04 | Improved |
| add distance correlation loss against a reference network | EfficientNet-B0 on ImageNet · PGD eps=0.1 transfer attack | accuracy (%): 11.17 -> 17.62 | Improved |
| add distance correlation loss against a reference network | ResNet-34 on ImageNet · PGD eps=0.1 transfer attack | accuracy (%): 41.70 -> 43.35 | Improved |
| add distance correlation loss against a reference network | ResNet-152 on ImageNet · PGD eps=0.1 transfer attack | accuracy (%): 44.49 -> 47.17 | Improved |
| add distance correlation loss against a reference network | ResNet-18 on CIFAR10 · clean accuracy | accuracy (%): 89.14 -> 87.61 | Hurt |
| add distance correlation loss against a reference network | MobileNet-v3-small on ImageNet · clean accuracy | accuracy (%): 47.16 -> 42.34 | Hurt |
| add distance correlation loss against a reference network | EfficientNet-B0 on ImageNet · clean accuracy | accuracy (%): 57.85 -> 55.82 | Hurt |
| add distance correlation loss against a reference network | ResNet-152 on ImageNet · clean accuracy | accuracy (%): 66.88 -> 68.04 | Improved |
| add distance correlation loss against a reference network | ResNet-34 on ImageNet · clean accuracy | accuracy (%): 64.01 -> 63.77 | No clear effect |
| use deeper network (Resnet 152 instead of Resnet 50) | pretrained ImageNet classifiers · partial dCor with class-name embedding | R^2(X, GT): 0.036 -> 0.019 | Hurt |
Closest papers by meaning
- Grounding Representation Similarity Through Statistical Testing (NeurIPS 2021 · similarity 0.89)
- Tracing Representation Progression: Analyzing and Enhancing Layer-Wise Similarity (ICLR 2025 · similarity 0.87)
- Estimating Shape Distances on Neural Representations with Limited Samples (ICLR 2024 · similarity 0.86)
- How Not to Stitch Representations to Measure Similarity: Task Loss Matching Versus Direct Matching (AAAI 2025 · similarity 0.85)
- Going Beyond Neural Network Feature Similarity: The Network Feature Complexity and Its Interpretation Using Category Theory (ICLR 2024 · similarity 0.85)
- Representational Similarity via Interpretable Visual Concepts (ICLR 2025 · 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.
