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

TLDR

Understanding and interpreting classification decisions of automated image classification systems is valuable for verifying reasoning and aiding experts, yet most machine learning methods act as black boxes that provide no insight into the factors driving decisions. This work proposes a general solution to this problem by decomposing nonlinear classifier decisions into pixel‑wise contributions. The authors present a methodology that visualizes single‑pixel contributions for kernel‑based Bag‑of‑Words classifiers and multilayered neural networks, and evaluate it on PASCAL VOC 2009, synthetic geometric shapes, MNIST, and a pre‑trained ImageNet model from Caffe. The resulting heatmaps enable human experts to verify classification validity and focus further analysis on regions of potential interest.

Abstract

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.