On Pixel-Wise Explanations for Non-Linear Classifier Decisions by Layer-Wise Relevance Propagation

Understanding and interpreting classification decisions of automated image classification systems is of high value in many applications, as it allows to verify the reasoning of the system and provides additional information to the human expert. Although machine learning methods are solving very succ...

Full description

Saved in:
Bibliographic Details
Published inPloS one Vol. 10; no. 7; p. e0130140
Main Authors Bach, Sebastian, Binder, Alexander, Montavon, Grégoire, Klauschen, Frederick, Müller, Klaus-Robert, Samek, Wojciech
Format Journal Article
LanguageEnglish
Published United States Public Library of Science 10.07.2015
Public Library of Science (PLoS)
Subjects
Algorithms
Artificial Intelligence
Artificial neural networks
Black boxes
Classification
Classification systems
Classifiers
Decision analysis
Decisions
Decomposition
Digits
Handwriting
Humans
Image classification
Image Processing, Computer-Assisted - methods
Information systems
Learning algorithms
Machine learning
Neural networks
Pattern Recognition, Automated - methods
Pixels
Taxonomy
VOCs
Volatile organic compounds
Online AccessGet full text

Cover

More Information
Summary:Understanding and interpreting classification decisions of automated image classification systems is of high value in many applications, as it allows to verify the reasoning of the system and provides additional information to the human expert. Although machine learning methods are solving very successfully a plethora of tasks, they have in most cases the disadvantage of acting as a black box, not providing any information about what made them arrive at a particular decision. This work proposes a general solution to the problem of understanding classification decisions by pixel-wise decomposition of nonlinear classifiers. We introduce a methodology that allows to visualize the contributions of single pixels to predictions for kernel-based classifiers over Bag of Words features and for multilayered neural networks. These pixel contributions can be visualized as heatmaps and are provided to a human expert who can intuitively not only verify the validity of the classification decision, but also focus further analysis on regions of potential interest. We evaluate our method for classifiers trained on PASCAL VOC 2009 images, synthetic image data containing geometric shapes, the MNIST handwritten digits data set and for the pre-trained ImageNet model available as part of the Caffe open source package.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 23
Competing Interests: The authors of this manuscript have read the journal’s policy and have the following competing interests: AB, FK, and KRM have a pending patent application: http://www.google.com/patents/WO2013037983A1?cl = en and http://www.google.com/patents/EP2570970A1?cl = en. This patent deals with pixel-wise visualization of bag of words features. This patent does not deal with artificial neural networks. The authors submitted a second patent application, “RELEVANCE SCORE ASSIGNMENT FOR ARTIFICIAL NEURAL NETWORKS” with the provisional application number PCT/EP2015/056008. The hereby confirm that this does not alter their adherence to PLOS ONE policies on sharing data and materials. All other authors (SB, GM, WS) have declared that no competing interests exist on their behalf.
Conceived and designed the experiments: SB AB GM WS FK KRM. Performed the experiments: SB AB GM. Wrote the paper: AB SB WS GM KRM FK. Conceived the theoretical framework: AB GM SB WS. Performed the experiments: SB AB GM. Revised the manuscript: KRM WS GM AB SB. Figure design: WS SB GM AB FK. Wrote code for MNIST: SB GM. Wrote code for ILSVRC: AB. Wrote code for BOW: SB AB. Image search and generation: AB SB GM.
ISSN:1932-6203
1932-6203
DOI:10.1371/journal.pone.0130140