Generation of Gridded Data for 2D Proportional Keen Approximator Using Radial Basis Functions

Abstract

A relatively young approximation method is the Proportional Keen Approximator (PKA). One of the main advantages of this method is its linear nature and consequently its low computational cost that makes it particularly suitable for real-time applications. However, the most significant drawback of this method is the restriction which it places on the form of training data. Strictly speaking, only gridded data can be used for training the PKA, implying the use of scattered training data is prohibited. In order to overcome this restriction, this paper considers the application of Radial Basis Functions (RBFs) for generation of gridded data from scattered ones. Also some experiments have been provided to further clarify the proposed method.