exhaustivePlot {gprege} | R Documentation |
Exhaustively searches the hyperparameter space by a grid, whose resolution is passed as an argument, and plots the LML function for every point in the space.
exhaustivePlot(y, x, xstar, options, maxwidth, res, nlevels)
y |
the target (output) data. |
x |
the input data matrix. |
xstar |
the points to predict function values. |
options |
options structure as defined by gpOptions.m. |
maxwidth |
maximum lengthscale to search for. |
res |
The search resolution. Number of points to plot for in the search range. |
nlevels |
Number of contour levels. |
area |
Area under the ROC curve of method-A. |
noiseLevel <- 0.2 noiseVar <- noiseLevel^2 options <- gpOptions() options$kern$comp <- list('rbf','white') ## Create data set l <- 9; x <- matrix(seq(0,240,by=20), ncol=1) trueKern <- kernCreate(x, 'rbf') trueKern$inverseWidth <- 1/(20^2) ## Characteristic inverse-width. K <- kernCompute(trueKern, x) + diag(dim(x)[1])*noiseVar ## Sample some true function values. y <- gaussSamp(Sigma=K, numSamps=1) xTest <- as.matrix(seq(0, 240, length=200)) graphics.off(); dev.new(); plot.new(); dev.new(); plot.new() exhaustivePlot(y, x, xTest, options=options, maxwidth=100, res=50, nlevels=75)