The training of the INeN is similar to that of the ISOM (Kurnaz et al., 2001). In the training
of the ISOM, the weights of the nodes were not modified; they were formed by the components
of the input feature vectors without modifying. In the training process of the INeN, the weights
of the nodes are modified according to Eq. (3). In each step of the training process, node weights
are updated by the input feature vector. In this way, the weights of the nodes represent statistical
distribution of the input feature vectors.