Comparative Analysis of Explainability Techniques for Non-linear Models

The aim of this final project is to inspect and compare dif- ferent techniques and tools available for Non-linear model explainability. The final objective is to explore the effectiveness of LIME [4], SHAP [1] and PSyKE [5] in returning comprehensive explanations for different types of machine learning algorithms; utilizing both images and tabular data. The effectiveness of these methods will be determined by assessing their consistency, comprehensibility and reliability.
The comparison will be made with respect to different algorithms, studying their ability to learn similar information.