A Generalized Chaotic Neural Network Model for Epilepsy Using Soboleva Hyperbolic Tangent Functions
Alexandra Choumpaev, Lazaros Moysis, Marcin Lawnik, George F. Fragulis · 2025
This work studies a generalized version of a chaotic neural network model for epilepsy. The model consists of several activation functions that are coupled to simulate the neuronal stimulation. The system is in a chaotic regime for normal behavior, and in a periodic regime for epilepsy. By replacing the hyperbolic tangent activation functions with the Soboleva hyperbolic tangent activation, the model is generalized through the introduction of four new control parameters for each neuron. The new model is studied through bifurcation and Lyapunov exponent diagrams, and reveals a collection of interesting dynamical phenomena. The use of the Soboleva function opens up a new direction towards the generalization of artificial neural networks for modeling neurological behaviors.