From attribution maps to human-understandable explanations through Concept Relevance Propagation

• Published in: Nature machine intelligence Vol. 5; no. 9; pp. 1006 - 1019
• Main Authors: Achtibat, Reduan, Dreyer, Maximilian, Eisenbraun, Ilona, Bosse, Sebastian, Wiegand, Thomas, Samek, Wojciech, Lapuschkin, Sebastian
• Format: Journal Article
• Language: English
• Published: London Nature Publishing Group UK 01.09.2023; Nature Publishing Group
• Subjects: 639/705/117; 639/705/531; Artificial intelligence; Back propagation; Decision making; Engineering; Explainable artificial intelligence; Neural networks; Neurons; Propagation; Reasoning; Semantics; Subspaces

The field of explainable artificial intelligence (XAI) aims to bring transparency to today’s powerful but opaque deep learning models. While local XAI methods explain individual predictions in the form of attribution maps, thereby identifying ‘where’ important features occur (but not providing infor...