saliency-cards
saliency-cards copied to clipboard
mitvis
Saliency Cards are transparency documentation for saliency methods. Learn about new saliency methods or document your own!
SHAP ([paper](https://proceedings.neurips.cc/paper_files/paper/2017/file/8a20a8621978632d76c43dfd28b67767-Paper.pdf)) is a commonly used and evaluated saliency method. Create a saliency card for it, including results from existing tests. Its methodology is characterized in Table A1 of the...
Guided backpropagation ([paper](https://arxiv.org/abs/1412.6806); [implementation](https://captum.ai/api/guided_backprop.html)) is a commonly used and evaluated saliency method. Create a saliency card for it, including results from existing tests. Its methodology is characterized in Table A1...
Existing perceptual correspondence evaluations have not characterized LIME. As a result, its [saliency card](https://github.com/mitvis/saliency-cards/blob/main/saliency_cards/lime/lime_saliency_card.md) does not have results for perceptual correspondence tests. Evaluate LIME's perceptual correspondence and record the results...
Existing model sensitivity evaluations have not characterized LIME. As a result, its [saliency card](https://github.com/mitvis/saliency-cards/blob/main/saliency_cards/lime/lime_saliency_card.md) does not have results for model sensitivity tests. Evaluate LIME's model sensitivity and record the results...
Existing minimality evaluations have not characterized LIME. As a result, its [saliency card](https://github.com/mitvis/saliency-cards/blob/main/saliency_cards/lime/lime_saliency_card.md) does not have results for minimality tests. Evaluate LIME's minimality and record the results in its saliency...
Existing label sensitivity evaluations have not characterized LIME. As a result, its [saliency card](https://github.com/mitvis/saliency-cards/blob/main/saliency_cards/lime/lime_saliency_card.md) does not have results for label sensitivity tests. Evaluate LIME's label sensitivity and record the results...
Existing input sensitivity evaluations have not characterized XRAI. As a result, its [saliency card](https://github.com/mitvis/saliency-cards/blob/main/saliency_cards/xrai/xrai_saliency_card.md) does not have results for an input sensitivity tests. Evaluate XRAI's input sensitivity and record the...
Existing label sensitivity evaluations have not characterized XRAI. As a result, its [saliency card](https://github.com/mitvis/saliency-cards/blob/main/saliency_cards/xrai/xrai_saliency_card.md) does not have results for label sensitivity tests. Evaluate XRAI's label sensitivity and record the results...
Existing minimality evaluations have not characterized XRAI. As a result, its [saliency card](https://github.com/mitvis/saliency-cards/blob/main/saliency_cards/xrai/xrai_saliency_card.md) does not have results for minimality tests. Evaluate XRAI's minimality and record the results in its saliency...