Research Article
REVEL Framework to Measure Local Linear Explanations for Black-Box Models: Deep Learning Image Classification Case Study
Table 56
Wilcoxon test for the different XAI methods and the prescriptivity metric in the CIFAR100 dataset.
| | LIME 2 | LIME 3 | LIME 4 | LIME 5 | LIME 6 | LIME 7 | LIME 8 | SHAP global | SHAP local |
| LIME 2 | — | — | — | — | — | — | — | — | — | LIME 3 | 0.241746 | — | — | — | — | — | — | — | — | LIME 4 | 0.543763 | 0.429915 | — | — | — | — | — | — | — | LIME 5 | 0.72744 | 0.90477 | 0.63844 | — | — | — | — | — | — | LIME 6 | 0.310117 | 0.944617 | 0.514841 | 0.780728 | — | — | — | — | — | LIME 7 | 0.629185 | 0.637106 | 0.808105 | 0.765539 | 0.626447 | — | — | — | — | LIME 8 | 0.227314 | 0.655869 | 0.2653 | 0.301112 | 0.598209 | 0.290457 | — | — | — | SHAP global | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | — | — | SHAP local | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | 0.0 | — |
|
|