Figure 4
From: Identification of subject-specific responses to footwear during running
Input relevance evaluation of the machine learning models (model explanations) trained for the. employed classification tasks: (A) subject x footwear, (B) subject, (C) footwear. The input relevance scores were obtained by Layer-wise Relevance Propagation (LRP). For each subfigure (A–C): The top part on the left shows the summed contribution of the relevance scores for each of the 101 time points of the stance phase. In the bottom part on the left, lighter colours indicate variables of high relevance, while darker colours indicate variables of low relevance. The bottom right part highlights the summed contribution of relevance scores of each of the ground reaction forces (GRFs), namely mediolateral (GRFML), anteroposterior (GRFAP), and vertical (GRFV). The figure was created using the MATLAB code provided by Hoitz et al. (2021)18.
