Supervised Contractive Neural Network
This ANNs calculates two net inputs for each node: a classic weighted input and a contractive input. This second net input tends to decay or to increase when the positive or negative value of the weight (w) becomes close to a specific constant (C). The contractive input works as a harmonic modulation of the weighted input.
References
[1] P.M. Buscema, G. Massini, G. Maurelli
Artificial Adaptive Systems to predict the magnitude of earthquakes
Bollettino di Geofisica Teorica ed Applicata Vol. 56, n. 2, pp. 227-256; June 2015.
[2] P.M. Buscema, R. Benzi
Quakes Prediction Using Highly Non Linear Systems and A Minimal Dataset
Chapter 4, pp 41-66, in P.M.Buscema & M.Ruggieri (eds), Advanced Networks, Algorithms and Modeling for Earthquake Prediction, River Publishers, 2011.