RBF (Radial Basis Function)

Part 2: RBF Consider a RBF network of n inputs (x1, x2, xs] and m outputs y = Di , Spose the outputs are weighted sum of RBF functions: 1-1 R,(x)= 20 where is a vector of dimension n, and H is the number of RFB hidden neurons. Derive the gradient of training erTor with respessary variables in the RBF network in order to use gradient-based training algorithms, assuming the training error per sample of data as where , represents one sample of the training data for the jth output