Explaining nonlinear classification decisions with deep Taylor decomposition

详细信息    查看全文

• 作者：Gr&eacute ; goire Montavon^a ; ^{gregoire.montavon@tu-berlin.de}Author Vitae ; Sebastian Lapuschkin^b ; ^{sebastian.lapuschkin@hhi.fraunhofer.de}Author Vitae ; Alexander Binder^c ; ^{alexander_binder@sutd.edu.sg}Author Vitae ; Wojciech Samek^b ; ^{wojciech.samek@hhi.fraunhofer.de}Author Vitae ; Klaus-Robert Mü ; ller^a ; ^d ; ^{klaus-robert.mueller@tu-berlin.de}Author Vitae
• 关键词：Deep neural networks ; Heatmapping ; Taylor decomposition ; Relevance propagation ; Image recognition
• 刊名：Pattern Recognition
• 出版年：2017
• 出版时间：May 2017
• 年：2017
• 卷：65
• 期：Complete
• 页码：211-222
• 全文大小：1501 K
• 卷排序：65

文摘

A novel method to explain nonlinear classification decisions in terms of input variables is introduced. The method is based on Taylor expansions and decomposes the output of a deep neural network in terms of input variables. The resulting deep Taylor decomposition can be applied directly to existing neural networks without retraining. The method is tested on two large-scale neural networks for image classification: BVLC CaffeNet and GoogleNet.