Authors: Szymon Maksymiuk (MI2DataLab Faculty of Mathematics and Information Science WUT) Alicja Gosiewska
Across R and Python environments there are plenty of tools dedicated to eXplainable Artificial Intelligence (XAI). On the one hand, there is a high diversity of methods, on the other hand, methods such as LIME or SHAP are implemented in many different libraries. The source of the popularity and success of some packages is the accessibility and convenience of use. In this talk, we will present popular XAI frameworks both from R and compare them by flexibility, the variety of explanations they provide, and their performance. We will show the comparison of XAI packages to familiarise the audience with the available solutions. Moreover, we will explore different classifications of the explanation methods, i.e. local vs global and model-agnostic vs model-specific explanations. The libraries covered in this talk are DALEXtra, flashlight, modelStudio, triplot and vip.