A computational design based scale conjugate neural network to solve the nonlinear Rabinovich–Fabrikant model

Purpose: The current research provides a design of computational framework using the scale conjugate gradient neural network for solving a nonlinear Rabinovich-Fabrikant model. The mathematical differential form of the nonlinear model has three classes p(x), q(x) and r(x), which is solved by applying the stochastic scale conjugate gradient neural network. Method: An Adam approach is used to get the dataset in order to reduce the mean square error with the division of the data 70% for training, while 15%, 15% for authorization and testing. Seventeen number of neurons, an activation log-sigmoid function, and a single layer feed-forward neural network is proposed to solve the Rabinovich-Fabrikant model. Results: By comparing the generated outputs with reference (Runge-Kutta) results, one can determine the validity of the proposed neural network. Furthermore, the precision of the designed neural computing network is judged by applying different performances of error histogram, correlation, and state transition. Novelty: The supervised scale conjugate gradient neural network has never been exploited before for solving the nonlinear Rabinovich-Fabrikant model.