inferAlpha: Infer alpha

In Caleb-Huo/MIS-Kmeans:

Description

MISKmenas

Usage

inferAlpha(resMISKmeans, balance = 1)

Arguments

resMISKmeans	The result returned from MISKmeans.
balance	selecting alpha such that the l1 norm penalty = balance * group penalty

Details

Estimate a propriate tuning parameter alpha

Value

alpha values

Author(s)

Caleb

Examples

S <- 2
K <- 3
G <- 1000
g1 <- 50
g2 <- 50
n0 <- 20
n <- K*n0
labels <- cut(1:n,breaks=K,labels=FALSE)

set.seed(32611)
S1 <- matrix(rnorm(G*n), nrow=G, ncol=n)
S2 <- matrix(rnorm(G*n), nrow=G, ncol=n)

S1[1:g1, labels==1] <- S1[1:g1, labels==1] + 2
S1[1:g1, labels==3] <- S1[1:g1, labels==3] - 2
S1[g1 + 1:g2, labels==1] <- S1[g1 + 1:g2, labels==1] - 2
S1[g1 + 1:g2, labels==2] <- S1[g1 + 1:g2, labels==2] + 2

S2[1:g1, labels==2] <- S2[1:g1, labels==2] + 2
S2[1:g1, labels==1] <- S2[1:g1, labels==1] - 2
S2[g1 + 1:g2, labels==2] <- S2[g1 + 1:g2, labels==2] - 2
S2[g1 + 1:g2, labels==3] <- S2[g1 + 1:g2, labels==3] + 2

S = list(t(S1),t(S2))
groups <- Map('c',1:g1,g1 + 1:g2)

res <- MISKmeans(d = S, K = 3, gamma = 0.4, group = groups)

inferAlpha(res) 
	

Caleb-Huo/MIS-Kmeans documentation built on May 17, 2019, 2:45 p.m.