Relative Layer-Wise Relevance Propagation: a more Robust Neural Networks eXplaination
This work addresses a specific technical bottleneck in explainable AI for neural networks, making it incremental but potentially useful for researchers and practitioners needing more robust explanations.
The authors tackled the problem of division by small values in Layer-Wise Relevance Propagation (LRP) for explainable machine learning by introducing Relative LRP, which satisfies the conservation law up to a multiplicative factor and eliminates this issue, except in ResNet skip connections, without requiring hyperparameter tuning.
Machine learning methods are solving very successfully a plethora of tasks, but they have the disadvantage of not providing any information about their decision. Consequently, estimating the reasoning of the system provides additional information. For this, Layer-Wise Relevance Propagation (LRP) is one of the methods in eXplainable Machine Learning (XML). Its purpose is to provide contributions of any neural network output in the domain of its input. The main drawback of current methods is mainly due to division by small values. To overcome this problem, we provide a new definition called Relative LRP where the classical conservation law is satisfied up to a multiplicative factor but without divisions by small values except for Resnet skip connection. In this article, we will focus on image classification. This allows us to visualize the contributions of a pixel to the predictions of a multi-layer neural network. Pixel contributions provide a focus to further analysis on regions of potential interest. R-LRP can be applied for any dense, CNN or residual neural networks. Moreover, R-LRP doesn't need any hyperparameters to tune contrary to other LRP methods. We then compare the R-LRP method on different datasets with simple CNN, VGG16, VGG19 and Resnet50 networks.