SFDesign (version 0.1.3)

maximin.augment: Augment a design by adding new design points that minimize the reciprocal distance criterion greedily

Description

This function augments a design by adding new design points one-at-a-time that minimize the reciprocal distance criterion.

Usage

maximin.augment(n, p, D.ini, candidate = NULL, r = 2 * p)

Value

the augmented design.

Arguments

n

the size of the final design.

p

design dimension.

D.ini

initial design.

candidate

candidate points to choose from. The default candidates are Sobol points of size 100n.

r

the power parameter in the maximin.crit. By default it is set as 2p.

Details

maximin.augment augments a design by adding new design points that minimize the reciprocal distance criterion (see maximin.crit) greedily. In each iteration, the new design points is selected as the one from the candidate points that has the smallest sum of reciprocal distance to the existing design, that is, \(\bm x_{k+1} = \arg\min_{\bm x} \sum_{i=1}^k \frac{1}{\|\bm x - \bm x_i\|^r}\).

Examples

n.ini = 10
n = 20
p = 3
D.ini = maximinLHD(n.ini, p)$design
D = maximin.augment(n, p, D.ini)