SOE: Soft Ordinal Embedding (SOE)

In loe: Local Ordinal Embedding

Description

Performs SOE for a given part of ordinal informaiton.

Usage

SOE(CM, N, p = 2, c = 0.1, maxit = 1000, report = 100, iniX = "rand",rnd=10000)

Arguments

CM	A matrix whose rows represent ordinal information: A row vector (i,j,k,l) means xi_{ij} < xi_{kl}, where xi_{ij} is a dissimilarity between object i and object j. If the number of given ordinal comparisons is more than 100000, then SOE randomly chooses rnd ordinal comarisons from given comparisons.
N	The number of objects.
p	The number of dimensions.
c	Scale parameter which only takes strictly positive value.
maxit	Maximum number of iteretions.
report	The frequency of reports. Defaults to every 100 iterations.
iniX	Matrix with starting values for embedding (optional).
rnd	The number of ordinal comarisons which are randomly chosen from given ordinal comparisons CM. rnd is used only if the number of given ordinal comparisons is more than 100000.

Value

SOE returns a list with components:

X	The best corrdinate matrix with p columns whose rows give the coordinates of objects.
str	The value of the stress function of SOE corresponding to X.

Author(s)

Yoshikazu Terada

Examples

library(MASS)
data(eurodist)
OIF <- get.order(as.matrix(eurodist))
sid <- sample(1:nrow(OIF))
POI <- OIF[sid[1:1000],]
result <- SOE(CM=POI, N=nrow(as.matrix(eurodist)),p=2, c=0.1,maxit =1000,report=100)
plot(result$X,type="n",xlab="Dim. 1", ylab="Dim. 2")
text(result$X,rownames(as.matrix(eurodist)),cex=1)

Example output

Loading required package: MASS
initial  value 879.590776 
final  value 0.996226 
converged

loe documentation built on May 2, 2019, 3:28 a.m.