designU: Select design in U

In mbinois/RRembo: Random EMbedding Bayesian Optimization for R

Select design in U

Description

Select design in U

Usage

designU(p, A, bxsize, type = "maximin", standard = FALSE)

Arguments

p	number of points of the design
A	random embedding matrix
bxsize	scaler bounds on the domain, i.e., bxsize x [-1,1]^D
type	design type, one of "LHS", "maximin" and 'unif'
standard	standard REMBO approach

Author(s)

Mickael Binois

References

M. Binois, D. Ginsbourger, O. Roustant (2018), On the choice of the low-dimensional domain for global optimization via random embeddings, arXiv:1704.05318

M. Binois (2015), Uncertainty quantification on Pareto fronts and high-dimensional strategies in Bayesian optimization, with applications in multi-objective automotive design, PhD thesis, Mines Saint-Etienne.

Examples

## Example of designs in U
set.seed(42)
d <- 2; D <- 5

A <- selectA(d, D, type = 'optimized')
size <- 5 # box size of Y
ntest <- 10000
Y <- size * (2 * matrix(runif(ntest * d), ntest, d) - 1)

inU <- testU(Y, A)
colors <- rep('black', ntest)
colors[inU] <- 'green'
plot(Y, col = colors, pch = 20, cex = 0.5)

p <- 20
designs1 <- designU(p, A, size)
designs2 <- designU(p, A, size, type = 'LHS')

points(designs1, col = 'red', pch = 20)
points(designs2, col = 'blue', pch = 20)
legend("topright", legend = c("LHS", "Maximin LHS"), col = c("blue", "red"), pch = c(20,20))

mbinois/RRembo documentation built on Sept. 16, 2023, 10:15 p.m.