Solution of Partial Differential Equations on Radial Basis Functions Networks

EasyChair Preprint no. 1964

Abstract

The solution of boundary value problems described by partial differential equations on networks of radial basis functions is considered. An analysis of gradient learning algorithms for radial basis functions networks showed that the widely used first-order method, the gradient descent method, does not provide a high learning speed and solution accuracy. The fastest method of the second order - the trust region method is very complex. A learning algorithm based on the Levenberg-Marquardt method is proposed. The proposed algorithm, with a simpler implementation, showed comparable results in comparison with the trust region method.

Keyphrases: Levenberg Marquardt method., neural network learning., Partial differential equations.

@Booklet{EasyChair:1964,
  author = {Mohie Alqezweeni and Vladimir Gorbachenko},
  title = {Solution of Partial Differential Equations on Radial Basis Functions Networks},
  howpublished = {EasyChair Preprint no. 1964},

  year = {EasyChair, 2019}}