A Neuron Sliding over a Boolean Matrix and Mathematical Prediction of Strong Earthquakes

摘要

The problem of predicting strong earthquakes, which cannot be solved by the methods of seismology, is solved by simple procedures of linear algebra and pattern recognition. For linear seismic regions (such as Kamchatka and the Kuril Islands), the prediction error determined by cross validation is at most two days, and the spatial accuracy of prediction is commensurate with the size of the seismic center; the probability of detection of strong earthquakes is about 0.8, and there are no false alarms.