AI Paper Atlas

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

A Benchmark for Interpretability Methods in Deep Neural Networks

Sara Hooker, Dumitru Erhan, Pieter-Jan Kindermans, Been Kim

NeurIPS 2019 · Label-efficient and robust learning › Recognition and long-tail · approach: Self-explaining and visual explanation networks · 389 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
Use base Gradients, Integrated Gradients, or Guided Backprop rankings instead of random rankingsfeature importance evaluation with ROAR · ImageNet, Birdsnap, Food 101accuracyHurt
Use Classic SmoothGrad averaging rather than a single estimateGradients and Integrated Gradients · ImageNet, Birdsnap, Food 101accuracyHurt
Use Classic SmoothGrad averaging rather than a single estimateGuided Backprop · ImageNet, Birdsnap, Food 101accuracyNo clear effect
Add squared noisy-estimate aggregation (SmoothGrad-Squared)base feature importance estimators · ImageNet, Birdsnap, Food 101accuracyImproved
Aggregate noisy estimates by variance (VarGrad)base feature importance estimators · ImageNet, Birdsnap, Food 101accuracyImproved
Square a single estimate before ranking, without ensembling (ablation)Gradients, Integrated Gradients, and Guided Backprop · ImageNet, Birdsnap, Food 101accuracy: 8.43% ± 0.97Improved
Replace removal with keeping the most important pixels (KAR) (ablation)feature importance estimators · ImageNet, Birdsnap, Food 101accuracyNo clear effect
Retrain on the modified training data rather than score modified inputs without retraining (ablation)ROAR evaluation · ImageNetaccuracy: 0.5% versus 63.53%Improved
Add VarGrad aggregation rather than using a single estimate (ablation)Guided Backprop · Birdsnap, ROAR 90.0% removedaccuracy: 59.76 -> 5.20Improved
Add SmoothGrad-Squared aggregation rather than using a single estimate (ablation)Guided Backprop · Birdsnap, ROAR 90.0% removedaccuracy: 59.76 -> 5.21Improved
Add SmoothGrad-Squared aggregation rather than using a single estimate (ablation)Gradients · Food 101, ROAR 90.0% removedaccuracy: 78.36 -> 8.34Improved
Add VarGrad aggregation rather than using a single estimate (ablation)Gradients · ImageNet, ROAR 90.0% removedaccuracy: 66.75 -> 10.41Improved

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