MATLAB: Sir what is the advantages of using RBF network for classification than Artificial neural network???In many clasiffication case RBF network is most commonly select than ANN??

Question

Sir what is the advantages of using RBF network for classification than Artificial neural network???In many clasiffication case RBF network is most commonly select than ANN???

Answer 1

If the graphical representation of the data indicates that classes are readily described by the superposition of elliptical and/or spherical shaped clusters, then elliptical and/or spherical shaped hidden neurons may be a better choice than sigmoidal hidden neurons.

Since both configurations are universal approximators, sometimes it just boils down to what makes the designer more comfortable.

As far as MATLAB software is concerned, the radial basis function program NEWRB is at a disadvantage because it does not allow an initial hidden neuron configuration to be specified. Conversely, the radial basis function program NEWRBE is at a worse disadvantage because it does not allow an initial hidden neuron configuration to be trained.

It is very sad that MATLAB has not made this improvement since I first mentioned it back in 2004.

Unfortunately, a version that I wrote in 2004 was lost in a computer crash.