Layer-Wise Relevance Propagation with Conservation Property for ResNet

TL;DR

This work extends Layer-Wise Relevance Propagation (LRP) to ResNet architectures by introducing Relevance Splitting at skip-connection convergence points, ensuring the conservation property for faithful explanations. It proposes Ratio-Based Splitting (and Symmetric alternatives) and Heat Quantization to produce sharper, more object-focused attribution maps while propagating relevance through both identity and projection skip connections. Empirical results on ImageNet and CUB demonstrate superior Insertion-Deletion (ID) scores and improved visual explanations compared with standard baselines, with ablations validating the design choices. The method offers a principled, conservation-guaranteed approach to interpreting residual networks and has potential applicability to other architectures with residual paths and beyond image modalities.

Abstract

The transparent formulation of explanation methods is essential for elucidating the predictions of neural networks, which are typically black-box models. Layer-wise Relevance Propagation (LRP) is a well-established method that transparently traces the flow of a model's prediction backward through its architecture by backpropagating relevance scores. However, the conventional LRP does not fully consider the existence of skip connections, and thus its application to the widely used ResNet architecture has not been thoroughly explored. In this study, we extend LRP to ResNet models by introducing Relevance Splitting at points where the output from a skip connection converges with that from a residual block. Our formulation guarantees the conservation property throughout the process, thereby preserving the integrity of the generated explanations. To evaluate the effectiveness of our approach, we conduct experiments on ImageNet and the Caltech-UCSD Birds-200-2011 dataset. Our method achieves superior performance to that of baseline methods on standard evaluation metrics such as the Insertion-Deletion score while maintaining its conservation property. We will release our code for further research at https://5ei74r0.github.io/lrp-for-resnet.page/

Figures (13)

• Figure 1: We propose LRP for ResNet. By formulating Relevance Splitting at a point where the output from a skip connection converges with that from a residual block, we extend LRP---originally designed for propagating relevance between two consecutive layers---to the ResNet architecture while guaranteeing its conservation property, thereby preserving the integrity of the explanation process.
• Figure 2: LRP for ResNet. Top: LRP propagates the relevance score backward to generate an attribution map corresponding to the input image. We focus on the relevance score propagation through the Bottleneck module, which incorporates a residual connection. Bottom: Architecture of the Bottleneck module. The D-Bottleneck employs a linear projection, in its skip connection for dimension matching. ReLU activation functions and batch normalization layers are omitted for simplicity.
• Figure 3: Left: Typical sample of an input image from ImageNet deng2009Imagenet. Right: the corresponding attribution as a visual explanation.
• Figure 4: Architecture of the Bottleneck block in ResNet and our Relevance Splitting approach. We introduce Relevance Splitting to consider the existence of skip connections in the relevance propagation of LRP.
• Figure 5: Qualitative Results: Attribution produced by each explanation method for the prediction of ResNet50 with respect to the ground-truth classes (top to bottom): "Brandt Cormorant," "Savannah Sparrow," "Sock," "Bustard," and "Bee." IG and Guided BP denote Integrated Gradients and Guided BackPropagation, respectively.