Visualization of radial basis function networks

Presents a method for the 3D visualization of the structure of radial basis function networks. This method allows the visualization of basis function characteristics (centers and widths) along with second level weights. Network properties can be displayed simultaneously with the training data or test data in the same input space. Principal component analysis is used to transform the input data so that its most salient dimensions can be visualized. This method also allows changes made while graphically editing the network structure, in transformed space, to be projected back into the original input space