demAutoOptimiseGp: Gaussian Process Optimisation Demo

In alkalait/gptk: Gaussian Processes Tool-Kit

Description

Shows that by varying the length scale, an artificial data set has different likelihoods, yet there is an optimum for which the likelihood is maximised. This demo is similar to demOptimiseGp, only here, it is demonstrated how the length scale hyperparameter is optimised automatically through SCG (scaled conjugate gradients) numerical optimisation. Run multiple times to understand the effect of optimisation on randomly generated datasets.

Usage

  demAutoOptimiseGp(path=getwd(), filename='demAutoOptimiseGp', png=FALSE, gif=FALSE)

Arguments

path	path where the plot images are saved.
filename	name of saved images.
png	save image as png.
gif	save series of images as animated gif.

See Also

gpOptions, kernCreate, gaussSamp, gpCreate, gpExpandParam, gpLogLikelihood, gpPosteriorMeanVar, gpOptimise, gpPlot.

alkalait/gptk documentation built on March 7, 2020, 6:30 a.m.