TY - JOUR

T1 - Q-SENN

T2 - 38th AAAI Conference on Artificial Intelligence, AAAI 2024

AU - Norrenbrock, Thomas

AU - Rudolph, Marco

AU - Rosenhahn, Bodo

N1 - Funding Information: This work was supported by the Federal Ministry of Education and Research (BMBF), Germany under the AI service center KISSKI (grant no. 01IS22093C) and the Deutsche Forschungsgemeinschaft (DFG) under Germany’s Excellence Strategy within the Cluster of Excellence PhoenixD (EXC 2122). This work was partially supported by Intel Corporation and by the German Federal Ministry of the Environment, Nature Conservation, Nuclear Safety and Consumer Protection (GreenAutoML4FAS project no. 67KI32007A).

PY - 2024/3/24

Y1 - 2024/3/24

N2 - Explanations in Computer Vision are often desired, but most Deep Neural Networks can only provide saliency maps with questionable faithfulness. Self-Explaining Neural Networks (SENN) extract interpretable concepts with fidelity, diversity, and grounding to combine them linearly for decision-making. While they can explain what was recognized, initial realizations lack accuracy and general applicability. We propose the Quantized-Self-Explaining Neural Network “Q-SENN”. Q-SENN satisfies or exceeds the desiderata of SENN while being applicable to more complex datasets and maintaining most or all of the accuracy of an uninterpretable baseline model, outperforming previous work in all considered metrics. Q-SENN describes the relationship between every class and feature as either positive, negative or neutral instead of an arbitrary number of possible relations, enforcing more binary human-friendly features. Since every class is assigned just 5 interpretable features on average, Q-SENN shows convincing local and global interpretability. Additionally, we propose a feature alignment method, capable of aligning learned features with human language-based concepts without additional supervision. Thus, what is learned can be more easily verbalized. The code is published: https://github.com/ThomasNorr/Q-SENN.

AB - Explanations in Computer Vision are often desired, but most Deep Neural Networks can only provide saliency maps with questionable faithfulness. Self-Explaining Neural Networks (SENN) extract interpretable concepts with fidelity, diversity, and grounding to combine them linearly for decision-making. While they can explain what was recognized, initial realizations lack accuracy and general applicability. We propose the Quantized-Self-Explaining Neural Network “Q-SENN”. Q-SENN satisfies or exceeds the desiderata of SENN while being applicable to more complex datasets and maintaining most or all of the accuracy of an uninterpretable baseline model, outperforming previous work in all considered metrics. Q-SENN describes the relationship between every class and feature as either positive, negative or neutral instead of an arbitrary number of possible relations, enforcing more binary human-friendly features. Since every class is assigned just 5 interpretable features on average, Q-SENN shows convincing local and global interpretability. Additionally, we propose a feature alignment method, capable of aligning learned features with human language-based concepts without additional supervision. Thus, what is learned can be more easily verbalized. The code is published: https://github.com/ThomasNorr/Q-SENN.

UR - http://www.scopus.com/inward/record.url?scp=85189611121&partnerID=8YFLogxK

U2 - 10.48550/arXiv.2312.13839

DO - 10.48550/arXiv.2312.13839

M3 - Conference article

AN - SCOPUS:85189611121

VL - 38

SP - 21482

EP - 21491

JO - Proceedings of the AAAI Conference on Artificial Intelligence

JF - Proceedings of the AAAI Conference on Artificial Intelligence

SN - 2159-5399

IS - 19

Y2 - 20 February 2024 through 27 February 2024

ER -