Unifying Attribution-Based Explanations Using Functional Decomposition

Arne Gevaert, Yvan Saeys

The black box problem in machine learning has led to the introduction of an ever-increasing set of explanation methods for complex models. These explanations have different propert…