Improving neural network interpretability via rule extraction
2018
Résumé
We present a method to replace the fully-connected layers of a Convolutional Neural Network (CNN9 with a small set of rules, allowing for better interpretation of its decisions while preserving accuracy.
Détails
Titre
Improving neural network interpretability via rule extraction
Auteur(s)/ trice(s)
Gomez Schnyder, Stéphane (School of Management and Engineering Vaud, HES-SO University of Applied Sciences Western Switzerland ; Computational Intelligence for Computational Biology (CI4CB), SIB Swiss Institute of Bioinformatics, Lausanne, Switzerland)
Despraz, Jérémie (School of Management and Engineering Vaud, HES-SO University of Applied Sciences Western Switzerland ; Computational Intelligence for Computational Biology (CI4CB), SIB Swiss Institute of Bioinformatics, Lausanne, Switzerland)
Peña-Reyes, Carlos Andrés (School of Management and Engineering Vaud, HES-SO University of Applied Sciences Western Switzerland ; Computational Intelligence for Computational Biology (CI4CB), SIB Swiss Institute of Bioinformatics, Lausanne, Switzerland)
Despraz, Jérémie (School of Management and Engineering Vaud, HES-SO University of Applied Sciences Western Switzerland ; Computational Intelligence for Computational Biology (CI4CB), SIB Swiss Institute of Bioinformatics, Lausanne, Switzerland)
Peña-Reyes, Carlos Andrés (School of Management and Engineering Vaud, HES-SO University of Applied Sciences Western Switzerland ; Computational Intelligence for Computational Biology (CI4CB), SIB Swiss Institute of Bioinformatics, Lausanne, Switzerland)
Date
2018-10
Publié dans
Proceedings Part I of Artificial Neural Networks and Machine Learning – ICANN 2018, 27th International Conference on Artificial Neural Networks, 4-7 October 2018, Rhodes, Greece
Volume
pp. 811-813
Editeur
Rhodes, Greece, 4-7 October 2018
Pagination
3 p.
Présenté à
Artificial Neural Networks and Machine Learning – ICANN 2018, 27th International Conference on Artificial Neural Networks, Rhodes, Greece, 2018-10-04, 2018-10-07
Mots-clés (libres)
convolutional neural network ; deep-learning rule extraction ; random forests ; interpretability
Type de papier
short paper
Domaine
Ingénierie et Architecture
Ecole
HEIG-VD
Institut
IICT - Institut des Technologies de l'Information et de la Communication
Le document apparaît dans
Documents de conférences
Global
Global