[4I1-GS-7] Language media processing:

[4I1-GS-7-03] Layer-Wise Relevance Propagation for ResNet: Visual Explanations Generation with Conservation Property

〇Seitaro Otsuki¹, Tsumugi Iida¹, Félix Doublet¹, Tsubasa Hirakawa², Takayoshi Yamashita², Hironobu Fujiyoshi², Komei Sugiura¹ (1. Keio University, 2. Chubu University)

Keywords:Explanable AI, Layer-Wise Relevance Propagation, ResNet

In the modern era, where deep learning is applied across a wide range of fields, the explainability of deep learning models is crucial, and the generation of explanations with high transparency is desirable. Layer-wise Relevance Propagation (LRP) is mentioned as a method for generating explanations by backpropagating relevance scores, offering high transparency. However, this method has been applied only to models without skip connections, as its application to models with skip connections does not satisfy the conservation property, resulting in poor quality explanations. Therefore, this paper proposes a method for calculating the backpropagation of relevance scores that satisfies the conservation property in models with skip connections. The proposed method outperformed existing methods in terms of Insertion-Deletion Score while satisfying the conservation property.