Description Usage Arguments Details Value Author(s) References Examples
Facilitates localized Gaussian process inference and prediction at a large
set of predictive locations, by opimizing a local set of inducing points
for each predictive location's local neighborhood and then calling giGP
.
1 2 3 |
XX |
a |
X |
a |
Y |
a vector of all responses/dependent values with |
M |
the positive integer number of inducing points placed for each local neighborhood; |
N |
the positive integer number of Nearest Neighbor (NN) locations used to build a local neighborhood |
g |
an initial setting or fixed value for the nugget parameter. In order to optimize g, a list can be provided that includes:
If |
theta |
an initial setting or fixed value for the lengthscale parameter. A (default)
If |
nu |
a positive number used to set the scale parameter;
default ( |
method |
specifies the method by which the inducing point template is built. In brief,
wIMSE ( |
integral_bounds |
a 2 by d |
num_thread |
a scalar positive integer indicating the number of threads to use for parallel processing |
epsK |
a small positive number added to the diagonal of the correlation matrix, of inducing points, K, for numerically stability for inversion. It is automatically increased if neccessary for each prediction. |
epsQ |
a small positive number added to the diagonal of the Q |
tol |
a positive number to serve as the tolerance level for covergence of the log-likelihood when optimizing the hyperparameter(s) theta and/or g |
reps |
a notification of replicate design locations in the data set. If |
This function builds a unique inducing point design to accompany the local neighborhood for each preditive location in XX
. It then invokes giGP
for each row of XX
with X=Xn, Y=Yn
from the corresponding local neighborhood and locally optimial inducing point design. For further information, see giGP
.
The output is a list
with the following components:
mean |
a vector of predictive means of length |
var |
a vector of predictive variances of length
|
nu |
a vector of values of the scale parameter of length
|
g |
a full version of the |
theta |
a full version of the |
Xm |
a |
eps |
a matrix of |
mle |
if |
time |
a scalar giving the passage of wall-clock time elapsed for (substantive parts of) the calculation |
D. Austin Cole austin.cole8@vt.edu
D.A. Cole, R.B. Christianson, and R.B. Gramacy (2021). Locally Induced Gaussian Processes for Large-Scale Simulation Experiments Statistics and Computing, 31(3), 1-21; preprint on arXiv:2008.12857; https://arxiv.org/abs/2008.12857
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 | library(hetGP); library(lhs)
X <- matrix(seq(0, 1, length=1000))
Y <- f1d(X)
XX <- matrix(seq(.01, .99, length=50))
YY <- f1d(XX)
n <- 50
m <- 7
int_bounds <- matrix(c(0,1))
out <- loiGP(XX=XX, X=X, Y=Y, M=m, N=n, method='wimse',
integral_bounds=int_bounds)
## Plot predicted mean and error
orig_par <- par()
par(mfrow=c(1,2))
plot(X, Y, type='l', lwd=4, ylim=c(-8, 16))
lines(XX, out$mean, lwd=3, lty=2, col=2)
legend('topleft', legend=c('Test Function', 'Predicted mean'),
lty=1:2, col=1:2, lwd=2)
plot(XX, YY - out$mean, xlab='X', ylab='Error', type = 'l')
par(orig_par)
|
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.