ON A LOCAL $k$-POINT MULTIQUADRIC NEURAL NETWORK IN FUNCTION RECOVERY APPLICATIONS

Abstract

To avoid problems caused by a global multiquadric network, a local manner is considered in this work. In a local influence domain, the number of evaluation points, represented by 'k ', is believed to play a crucial role in determining the success of the scheme. In this work, this 'k ' is numerically investigated under a local manner of MQ-RBF neural network when applied to function approximation and recovery applications. The results discovered in this work strongly indicate that with a good combination of an MQ format and a shape-parameter choosing strategy, an optimal interval of k can be located with high confidence. This piece of insight is certainly useful for further applications of the scheme.