Search of hidden periodicities in noisy symbolic sequences with neural networks

We describe the use of neural network for the search of hidden periodicities in the noisy symbolic sequences. The approach is based on the application of generalized Hopfield neural network. This network serves for the extraction of prototypes corresponding to subsequences obtained by the various 1-subdivisions of an initial sequence. The criterion for the statistical significance of the prototypes is given within the internal terms. The method is applicable in the case of superposition and simultaneous coexistence of different hidden periodicities.