For the subspace clustering, traditional method of least squares regression
is used to build coefficient matrix that reconstructs the data point by solving
$$\textrm{min}_Z \|X-XZ\|_F^2 + \lambda \|Z\|_F \textrm{ such that }diag(Z)=0$$
where \(X\in\mathbf{R}^{p\times n}\) is a column-stacked data matrix.
As seen from the equation, we use a denoising version controlled by \(\lambda\) and
provide an option to abide by the constraint \(diag(Z)=0\) by zerodiag parameter.
LSR(data, k = 2, lambda = 1e-05, zerodiag = TRUE)a named list of S3 class T4cluster containing
a length-\(n\) vector of class labels (from \(1:k\)).
name of the algorithm.
Lu C, Min H, Zhao Z, Zhu L, Huang D, Yan S (2012). “Robust and Efficient Subspace Segmentation via Least Squares Regression.” In Hutchison D, Kanade T, Kittler J, Kleinberg JM, Mattern F, Mitchell JC, Naor M, Nierstrasz O, Pandu Rangan C, Steffen B, Sudan M, Terzopoulos D, Tygar D, Vardi MY, Weikum G, Fitzgibbon A, Lazebnik S, Perona P, Sato Y, Schmid C (eds.), Computer Vision -ECCV 2012, volume 7578, 347–360. Springer Berlin Heidelberg, Berlin, Heidelberg. ISBN 978-3-642-33785-7 978-3-642-33786-4.
# \donttest{
## generate a toy example
set.seed(10)
tester = genLP(n=100, nl=2, np=1, iso.var=0.1)
data = tester$data
label = tester$class
## do PCA for data reduction
proj = base::eigen(stats::cov(data))$vectors[,1:2]
dat2 = data%*%proj
## run LSR for k=3 with different lambda values
out1 = LSR(data, k=3, lambda=1e-2)
out2 = LSR(data, k=3, lambda=1)
out3 = LSR(data, k=3, lambda=1e+2)
## extract label information
lab1 = out1$cluster
lab2 = out2$cluster
lab3 = out3$cluster
## visualize
opar <- par(no.readonly=TRUE)
par(mfrow=c(1,3))
plot(dat2, pch=19, cex=0.9, col=lab1, main="LSR:lambda=1e-2")
plot(dat2, pch=19, cex=0.9, col=lab2, main="LSR:lambda=1")
plot(dat2, pch=19, cex=0.9, col=lab3, main="LSR:lambda=1e+2")
par(opar)
# }