Numerical Solution of Stiff Ordinary Differential Equations with Random Projection Neural Networks

  • 2021-08-03 15:49:17
  • Evangelos Galaris, Francesco CalabrĂ², Daniela di Serafino, Constantinos Siettos
  • 1

Abstract

We propose a numerical scheme based on Random Projection Neural Networks(RPNN) for the solution of Ordinary Differential Equations (ODEs) with a focuson stiff problems. In particular, we use an Extreme Learning Machine, asingle-hidden layer Feedforward Neural Network with Radial Basis Functionswhich widths are uniformly distributed random variables, while the values ofthe weights between the input and the hidden layer are set equal to one. Thenumerical solution is obtained by constructing a system of nonlinear algebraicequations, which is solved with respect to the output weights using theGauss-Newton method. For our illustrations, we apply the proposed machinelearning approach to solve two benchmark stiff problems, namely the Rober andthe van der Pol ones (the latter with large values of the stiffness parameter),and we perform a comparison with well-established methods such as the adaptiveRunge-Kutta method based on the Dormand-Prince pair, and a variable-stepvariable-order multistep solver based on numerical differentiation formulas, asimplemented in the \texttt{ode45} and \texttt{ode15s} MATLAB functions,respectively. We show that our proposed scheme yields good numericalapproximation accuracy without being affected by the stiffness, thusoutperforming in same cases the \texttt{ode45} and \texttt{ode15s} functions.Importantly, upon training using a fixed number of collocation points, theproposed scheme approximates the solution in the whole domain in contrast tothe classical time integration methods.

 

Quick Read (beta)

loading the full paper ...