Numerical Comparison of Shapeless Radial Basis Function Networks in燩attern Recognition

This work focuses on radial basis functions containing no parameters with the main objective being to comparatively explore more of their effectiveness. For this, a total sixteen forms shapeless are gathered and investigated under context pattern recognition problem through structure function neural networks, use Representational Capability (RC) algorithm. Different sizes datasets disturbed noise before imported into algorithm as ‘training/testing’ datasets. Each is monitored carefully effectiveness criteria including accuracy, condition number (of interpolation matrix), CPU time, CPU-storage requirement, underfitting overfitting aspects, centres generated. sake comparison, well-known Multiquadric-radial included representative shape-contained functions. The numerical results have revealed that some show good potential even better than Multiquadric itself indicating strongly future may longer face pain choosing proper shape when be equally (or better) effective.