Daniel@0: Daniel@0:
Daniel@0:Daniel@0: demev3Daniel@0: Daniel@0: Daniel@0:
x
which sampled from a
Daniel@0: Gaussian distribution, and a target variable t
generated by
Daniel@0: computing sin(2*pi*x)
and adding Gaussian noise. An RBF
Daniel@0: network with linear outputs is trained by minimizing a sum-of-squares
Daniel@0: error function with isotropic Gaussian regularizer, using the scaled
Daniel@0: conjugate gradient optimizer. The hyperparameters alpha
and
Daniel@0: beta
are re-estimated using the function evidence
. A graph
Daniel@0: is plotted of the original function, the training data, the trained
Daniel@0: network function, and the error bars.
Daniel@0:
Daniel@0: demev1
, evidence
, rbf
, scg
, netevfwd
Copyright (c) Ian T Nabney (1996-9) Daniel@0: Daniel@0: Daniel@0: Daniel@0: