AI Paper Atlas

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

ECCV 2022 · Adversarial robustness and security › Certified and empirical robustness · approach: Robust overfitting and regularization · 30 citations (OpenAlex)

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 changedWhereMeasuredEffect
add distance correlation loss against a reference networkResNet-18 on CIFAR10 · PGD eps=0.1 transfer attackaccuracy (%): 27.41 -> 36.11Improved
add distance correlation loss against a reference networkResNet-18 on CIFAR10 · PGD eps=0.05 transfer attackaccuracy (%): 49.42 -> 59.33Improved
add distance correlation loss against a reference networkMobileNet-v3-small on ImageNet · PGD eps=0.1 transfer attackaccuracy (%): 17.15 -> 28.04Improved
add distance correlation loss against a reference networkEfficientNet-B0 on ImageNet · PGD eps=0.1 transfer attackaccuracy (%): 11.17 -> 17.62Improved
add distance correlation loss against a reference networkResNet-34 on ImageNet · PGD eps=0.1 transfer attackaccuracy (%): 41.70 -> 43.35Improved
add distance correlation loss against a reference networkResNet-152 on ImageNet · PGD eps=0.1 transfer attackaccuracy (%): 44.49 -> 47.17Improved
add distance correlation loss against a reference networkResNet-18 on CIFAR10 · clean accuracyaccuracy (%): 89.14 -> 87.61Hurt
add distance correlation loss against a reference networkMobileNet-v3-small on ImageNet · clean accuracyaccuracy (%): 47.16 -> 42.34Hurt
add distance correlation loss against a reference networkEfficientNet-B0 on ImageNet · clean accuracyaccuracy (%): 57.85 -> 55.82Hurt
add distance correlation loss against a reference networkResNet-152 on ImageNet · clean accuracyaccuracy (%): 66.88 -> 68.04Improved
add distance correlation loss against a reference networkResNet-34 on ImageNet · clean accuracyaccuracy (%): 64.01 -> 63.77No clear effect
use deeper network (Resnet 152 instead of Resnet 50)pretrained ImageNet classifiers · partial dCor with class-name embeddingR^2(X, GT): 0.036 -> 0.019Hurt

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy