On divided neural network and its application to source localization in the brain

Decomposition of an artificial neural network (ANN) into two or more component neural networks, called divided neural network (DNN), is considered. This decomposition is very useful for solving some inverse problems in which certain parameters in the output are more important than others in the system under consideration. It is shown that, in this case, the DNN performs superior to the ANN. As an example, the application of the DNN to dipole source localization from a 64-channel EEG signal illustrates the authors' theoretical result.