fastClustering: Fast Spectral Clustering

In sClust: R Toolbox for Unsupervised Spectral Clustering

Description

This function will sample the data before performing a classification function on the samples and then applying K nearest neighbours.

Usage

fastClustering(
  dataFrame,
  smplPoint,
  stopCriteria = 0.99,
  neighbours = 7,
  similarity = TRUE,
  clustFunction,
  ...
)

Arguments

dataFrame	The dataFrame.
smplPoint	maximum of sample number for reduction.
stopCriteria	criterion for minimizing intra-group distance and select final smplPoint.
neighbours	number of points that will be selected for the similarity computation.
similarity	if True, will use the similarity matrix for the clustering function.
clustFunction	the clustering function to apply on data.
...	additional arguments for the clustering function.

Value

returns a list containing the following elements:

• results: clustering results

• sample: dataframe containing the sample used

• quantLabels: quantization labels

• clustLabels: results labels

• kmeans: kmeans quantization results

Author(s)

Emilie Poisson Caillault and Erwan Vincent

Examples

### Example 1: 2 disks of the same size
n<-100 ; r1<-1
x<-(runif(n)-0.5)*2;
y<-(runif(n)-0.5)*2
keep1<-which((x*2+y*2)<(r1*2))
disk1<-data.frame(x+3*r1,y)[keep1,]
disk2 <-data.frame(x-3*r1,y)[keep1,]
sameTwoDisks <- rbind(disk1,disk2)
res <- fastClustering(scale(sameTwoDisks),smplPoint = 500, 
                      stopCriteria = 0.99, neighbours = 7, similarity = TRUE,
                      clustFunction = UnormalizedSC, K = 2)
plot(sameTwoDisks, col = as.factor(res$clustLabels))

### Example 2: Speed and Stopping Distances of Cars
res <- fastClustering(scale(iris[,-5]),smplPoint = 500, 
                      stopCriteria = 0.99, neighbours = 7, similarity = TRUE,
                      clustFunction = spectralPAM, K = 3)
plot(iris, col = as.factor(res$clustLabels))
table(res$clustLabels,iris$Species)

sClust documentation built on Aug. 24, 2021, 1:06 a.m.