expectedImprovement: Expected Improvement

In bartzbeielstein/SPOT: Sequential Parameter Optimization Toolbox

Description

Compute the negative logarithm of the Expected Improvement of a set of candidate solutions. Based on mean and standard deviation of a candidate solution, this estimates the expectation of improvement. Improvement considers the amount by which the best known value (best observed value) is exceeded by the candidates.

Usage

expectedImprovement(mean, sd, min)

Arguments

mean	vector of predicted means of the candidate solutions.
sd	vector of estimated uncertainties / standard deviations of the candidate solutions.
min	minimal observed value.

Value

a vector with the negative logarithm of the expected improvement values, -log10(EI).

Examples

mean <- 1:10 #mean of the candidates
sd <- 10:1 #st. deviation of the candidates
min <- 5 #best known value
EI <- expectedImprovement(mean,sd,min)
EI

bartzbeielstein/SPOT documentation built on June 13, 2020, 5:58 p.m.