SPECC: Performs Spectral Clustering

In alyssajs/CS599: Hierarchical and K-Means Clustering

Description

Performs spectral clustering using K-nearest neighbors (where K is passed in as a parameter) using Euclidean distances. Uses K-means clustering on the eigenvectors.

Usage

SPECC(data.dt, numClust, numEigen, numNeighbors)

Arguments

data.dt	Data table of observations
numClust	Desired number of clusters
numEigen	Desired number of eigenvalues used
numNeighbors	Number of neighbors to choose in K-nearest neighbor computation

Value

Vector of cluster assignments.

Examples

library(data.table)
set.seed(1)
halfcircle <- function(r, center = c(0, 0), class, sign, N=150, noise=0.5) {
 angle <- runif(N, 0, pi)
 rad <- rnorm(N, r, noise)
 data.table(
   V1 = rad * cos(angle) + center[1],
   V2 = sign * rad * sin(angle) + center[2]
   )
}
X.dt <- rbind(
 halfcircle(4, c(0, 0), 1, 1),
 halfcircle(4, c(4, 2), 2, -1))

result <- SPECC(X.dt, 2, 5, 5)

alyssajs/CS599 documentation built on June 17, 2021, 6:29 p.m.