Learning process of Affordable Neural Network for backpropagation algorithm

We have recently proposed a novel neural network structure called an “Affordable Neural Network” (AfNN), in which affordable neurons of the hidden layer are considered as the elements responsible for the robustness property as is observed in human brain function. We have confirmed that the AfNN gains good performance both of the generalization ability and the learning ability. Furthermore, the AfNN has durability, because the AfNN still performs well even if some of neurons in the hidden layer are damaged after learning process. In this study, we study the characteristics of weights of the AfNN during the learning process to make clear the reason of that the AfNNs can perform well for learning and generalization abilities and operate as usually against damaging neurons.