Example to use an external covariance function.
demogp_matern
- script to exemplify the use of an external covariance
function. The function chosen is the Matern covariance function which
allows the transition from a rough (once- or twice-differentiable) kernel to
the infinitely many times differentiable RBF kernel (implemented in this
package as 'sqexp'
).
The demonstration starts from sampling from a realisation of a GP with zero
mean and a matern
covariance of order 1.5
.
This sample function is corrupted with additive Gaussian noise, then a GP
with matern kernel - but different parameters - is trained on the samples.
In the demo script the experiments are repeated three times and different BV set sizes are also tested (to make it faster, change the parameters accordingly).
ogp
, ogptrain
, cov_matern
, sincdata
Copyright (c) Lehel Csató (2001-2004)