# RBF Network for classification

I would like to know how it is calculated the outcomes (i.e. the output layer output) of a RBF Network for a classification problem.

My code fits the hidden->output weights with linear regression and then uses the fitted model to predict. However, this doesn't seem right because in classification problems we have a discrete range of values as labels, and my network is providing continuous outcomes (real numbers).

For readability purposes, i pasted the full code here. Any guidance would be greatly helpful.