AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

“Why Not Other Classes?”: Towards Class-Contrastive Back-Propagation Explanations

Yipei Wang, Xiaoqian Wang

NeurIPS 2022 · Label-efficient and robust learning › Recognition and long-tail · approach: Self-explaining and visual explanation networks · 4 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
weighted contrast instead of original logit back-propagation, keep positive features (Gaussian Blur)GradCAM · CUB-200 t2expected relative probability: 0.406 -> 0.776Improved
weighted contrast instead of original, keep positive features (Gaussian Blur)Linear Approximation · CUB-200 t2expected relative probability: 0.450 -> 0.749Improved
weighted contrast instead of original, keep positive features (Gaussian Blur)GradCAM · Food-101 t2expected relative probability: 0.669 -> 0.888Improved
weighted contrast instead of original, keep negative features (Gaussian Blur)GradCAM · Food-101 t2expected relative probability: 0.317 -> 0.085Improved
weighted contrast instead of original, keep positive features (Gaussian Blur)GradCAM · Stanford Cars t1expected relative probability: 0.810 -> 0.899Improved
weighted contrast instead of original, keep positive features (Gaussian Blur)Linear Approximation · FGVC t1expected relative probability: 0.757 -> 0.808Improved
weighted contrast instead of original, keep positive features (Gaussian Blur)Linear Approximation · Flower-102 t2expected relative probability: 0.598 -> 0.746Improved
weighted contrast instead of gradient of logit, 1/2/10 PGD iterationsinput gradient · ILSVRC2012, 9 CNNstarget probability pt and accuracyImproved
weighted contrast instead of mean contrastGradCAM · ILSVRC2012 / CUB-200explanation qualityImproved
weighted contrast instead of max contrastGradCAM · CUB-200explanation qualityImproved
change blur baseline to zeros or channel-wise mean (ablation)weighted GradCAM · CUB-200 t2expected relative probability: 0.776 -> 0.762No clear effect
explain probability instead of logitback-propagation explanations · CUB-200localization vs contrastMixed

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