exam_train and exam_test are example scripts of how to train and
predict with the warped GP. Further information can be gained by
examining comments within the code. Files with 01 refer to Gaussian
covariance, and those with 03 refer to neural net covariance.

The plotwarp functions show the warping function learnt. For some
datasets the warping function is a straight line, in which case the
warped GP has not been able to learn anything extra. On the other
hand if the warping function looks non-linear then hopefully a useful
transformation has been learnt which improves on the standard GP.

All code is based on, and uses, Carl Edward Rasmussen's code for GPs
which is also supplied here: gp01lik, gp01pred, gp03lik, gp03pred,
minimize.