System identification of a circadian signal source using BP neural networks

The dynamics of some biological signal sources may be identified through measured data. The regulating characteristics of the wake-sleep circadian rhythm is identified here as an example by using BP neural networks. As a result, a MA model of neural networks can acquire the characteristics. The change of wake-sleep period is almost controlled suppressively by the data of preceding three days. The change of the dynamics can be evaluated by internal representation of the network. This method may be useful for medical diagnoses.