Skip to main content
. 2021 Aug 17;9:704738. doi: 10.3389/fbioe.2021.704738

FIGURE 4.

FIGURE 4

Illustration of layer-wise (backwards) relevance propagation in a CANN. In a trained neural network, recursive application of this scheme from the output layer to the input layer can quantify the relevance of the individual input parameters such as the ELISA values for the output. Neurons are illustrated as empty circles, weights as black arrows.