Who Cited It

Explaining nonlinear classification decisions with deep Taylor decomposition

2016 · Pattern Recognition · 1,413 citations · 9 from inside this corpus

Grégoire Montavon, Sebastian Lapuschkin, Alexander Binder, Wojciech Samek, Klaus‐Robert Müller

Nonlinear methods such as Deep Neural Networks (DNNs) are the gold standard for various challenging machine learning problems such as image recognition. Although these methods perform impressively well, they have a significant disadvantage, the lack of transparency, limiting the interpretability of the solution and thus the scope of application in practice. Especially DNNs act as black boxes due to their multilayer nonlinear structure. In this paper we introduce a novel methodology for interpreting generic multilayer neural networks by decomposing the network classification decision into contributions of its input elements. Although our focus is on image classification, the method is applicable to a broad set of input data, learning tasks and network architectures. Our method called deep Taylor decomposition efficiently utilizes the structure of the network by backpropagating the explanations from the output to the input layer. We evaluate the proposed method empirically on the MNIST and ILSVRC data sets.

Explaining nonlinear classification decisions with deep Taylor decomposition (2016)Explaining nonlinear classifi…Random Forests (2001)Random ForestsLearning representations by back-propagating errors (1986)Learning representations by b…Neural networks for pattern recognition (1994)Neural networks for pattern r…Neural Networks for Pattern Recognition (1995)Neural Networks for Pattern R…Gaussian Processes for Machine Learning (2005)Gaussian Processes for Machin…Learning with Kernels (2001)Learning with KernelsLearning Deep Architectures for AI (2009)Learning Deep Architectures f…Recursive Deep Models for Semantic Compositionality Over a Sentiment Treebank (2013)Recursive Deep Models for Sem…Intriguing properties of neural networks (2013)Intriguing properties of neur…DeepFool: A Simple and Accurate Method to Fool Deep Neural Networks (2016)DeepFool: A Simple and Accura…Natural Language Processing (almost) from Scratch (2011)Natural Language Processing (…Learning With Kernels: Support Vector Machines, Regularization, Optimization, and Beyond (2003)Learning With Kernels: Suppor…Natural Language Processing (almost) from Scratch (2011)Natural Language Processing (…An introduction to kernel-based learning algorithms (2001)An introduction to kernel-bas…Learning in Graphical Models (1998)Learning in Graphical ModelsNeural Networks: Tricks of the Trade (2012)Neural Networks: Tricks of th…Interpreting neural-network connection weights (1991)Interpreting neural-network c…A survey on deep learning in medical image analysis (2017)A survey on deep learning in …Explainable Artificial Intelligence (XAI): Concepts, taxonomies, opportunities and challe… (2019)Explainable Artificial Intell…Peeking Inside the Black-Box: A Survey on Explainable Artificial Intelligence (XAI) (2018)Peeking Inside the Black-Box:…A survey of methods for explaining black box models (2019)A survey of methods for expla…Explainable AI: A Review of Machine Learning Interpretability Methods (2020)Explainable AI: A Review of M…Methods for interpreting and understanding deep neural networks (2017)Methods for interpreting and …Machine Learning Interpretability: A Survey on Methods and Metrics (2019)Machine Learning Interpretabi…Explainable Artificial Intelligence (XAI): What we know and what is left to attain Trustw… (2023)Explaining Deep Neural Networks and Beyond: A Review of Methods and Applications (2021)Explaining Deep Neural Networ…
26 of 26 neighbouring works in this corpus. Blue is what this paper cites; orange is what cites it, and a dashed line is one neighbour citing another. Only the largest labels are drawn — every node carries its full title on hover.
this paper works it cites works citing it node size = global citations · hover for the full title

Citation edge evidence

Each row is directed from the citing work to the cited work. Source status remains readable without the interactive graph.

What this paper cites, inside the corpus

PaperYearCited
Random Forests2001131,109
Learning representations by back-propagating errors198631,688
Neural networks for pattern recognition199418,720
Neural Networks for Pattern Recognition199512,273
Gaussian Processes for Machine Learning200510,470
Learning with Kernels20019,631
Learning Deep Architectures for AI20096,951
Recursive Deep Models for Semantic Compositionality Over a Sentiment Treebank20136,849
Intriguing properties of neural networks20135,739
DeepFool: A Simple and Accurate Method to Fool Deep Neural Networks20165,392
Natural Language Processing (almost) from Scratch20115,174
Learning With Kernels: Support Vector Machines, Regularization, Optimization, and Beyond20034,332
Natural Language Processing (almost) from Scratch20113,991
An introduction to kernel-based learning algorithms20013,498
Learning in Graphical Models19982,437
Neural Networks: Tricks of the Trade20122,159
Interpreting neural-network connection weights19911,499

What cites it, inside the corpus

PaperYearCited
A survey on deep learning in medical image analysis201715,110
Explainable Artificial Intelligence (XAI): Concepts, taxonomies, opportunities and challe…20199,796
Peeking Inside the Black-Box: A Survey on Explainable Artificial Intelligence (XAI)20186,166
A survey of methods for explaining black box models20194,992
Explainable AI: A Review of Machine Learning Interpretability Methods20202,907
Methods for interpreting and understanding deep neural networks20172,763
Machine Learning Interpretability: A Survey on Methods and Metrics20191,822
Explainable Artificial Intelligence (XAI): What we know and what is left to attain Trustw…20231,679
Explaining Deep Neural Networks and Beyond: A Review of Methods and Applications20211,424

Links

Download BibTeX · Download RIS · DOI · Open access copy · OpenAlex record

Topics

Explainable Artificial Intelligence (XAI)Computer Science
Generative Adversarial Networks and Image SynthesisComputer Science
Model Reduction and Neural NetworksPhysics and Astronomy

Is this record sound?

complete

Nothing in this record contradicts itself and no field we check is missing.

  • supports5 author record(s) attached.
  • supports78 reference(s) recorded.
  • supportsThe DOI's year agrees with the publication year.
  • supportsA title is present.

Provenance

Everything above was read from one stored OpenAlex payload, fetched 2026-09-04T03:58:59+00:00.

sha256 59007f057c7dae86…