AI Paper Atlas

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

Statistically Consistent Saliency Estimation

Shunyan Luo, Emre Barut, Fang Jin

ICCV 2021 · Learning theory and algorithms › Prediction and decision losses · approach: Provable post-hoc explanations · 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
use LEG instead of LIMEsaliency estimation on VGG-19 · ImageNet (500 images), Distance-100 maskingaverage minimal perturbation size S_key: 0.269 -> 0.163Improved
use LEG instead of KernelSHAPsaliency estimation on VGG-19 · ImageNet (500 images), Distance-100 maskingaverage minimal perturbation size S_key: 0.251 -> 0.163Improved
use LEG instead of KernelSHAPsaliency estimation on VGG-19 · ImageNet (500 images), Mean-vector maskingaverage minimal perturbation size S_key: 0.074 -> 0.118Hurt
use LEG instead of C-Shapleysaliency estimation on VGG-19 · ImageNet (500 images), Black-out maskingaverage minimal perturbation size S_key: 0.101 -> 0.062Improved
add total-variation penalty (LEG-TV vs LEG) (ablation)LEG · ImageNet (500 images), Black-out maskingaverage minimal perturbation size S_key: 0.062 -> 0.056No clear effect
add total-variation penalty (LEG-TV vs LEG) (ablation)LEG · ImageNet (500 images), Distance-100 maskingaverage minimal perturbation size S_key: 0.163 -> 0.170No clear effect
use LEG-TV instead of GradCamsaliency estimation on VGG-19 · ImageNet (500 images), Distance-100 maskingaverage minimal perturbation size S_key: 0.318 -> 0.170Improved
use LEG instead of KernelSHAPsaliency estimation on VGG-19 · ImageNet (500 images), Noise maskingaverage minimal perturbation size S_key: 0.060 -> 0.069Hurt

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