AI Paper Atlas

Research map / Learning theory and algorithms / Prediction and decision losses

On the (In)fidelity and Sensitivity of Explanations

Chih-Kuan Yeh, Cheng-Yu Hsieh, Arun Suggala, David I Inouye, Pradeep K Ravikumar

NeurIPS 2019 · Learning theory and algorithms › Prediction and decision losses · approach: Provable post-hoc explanations

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
Apply SmoothGrad to Grad (ablation)Grad · MNIST globalsensitivity: 0.56 -> 0.28Improved
Apply SmoothGrad to Grad (ablation)Grad · MNIST globalinfidelity: 2.38 -> 1.89Improved
Apply SmoothGrad to base explanations (ablation)Grad, IG, and GBP · MNIST, CIFAR-10, and ImageNet globalsensitivity and infidelityImproved
Use the optimal explanation for square-removal perturbationsGlobal explanations · MNIST, CIFAR-10, and ImageNetinfidelityImproved
Use the optimal explanation for noisy-baseline perturbationsLocal explanations · MNISTinfidelityImproved
Apply the optimal infidelity explanationExplanations in the human evaluation · Human evaluationhuman accuracy: 0.53 -> 0.88Improved
Increase SmoothGrad smoothing radius (ablation)Grad-SG · MNIST localinfidelityMixed
Increase SmoothGrad smoothing radius (ablation)Grad-SG · MNIST localsensitivityImproved
Adversarially train the predictor (ablation)Gradient-based explanations · MNIST robust vs regular networksensitivity and infidelityImproved
Use a constant explanationAny explanation · All inputssensitivityImproved

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