Package 'ScorePlus'

find the maxinum distance from the convex hull formed by the chosen K vertices

Description

find the maxinum distance from the convex hull formed by the chosen K vertices

Usage

getMaxDist(centers, vertex.ind)

Arguments

centers	L-by-(K-1) center matrix
vertex.ind	index of the K centers that forms the convex hull

Value

the maximum distance

---

calculated the membership of each node given ratio matrix and community centers

Description

calculated the membership of each node given ratio matrix and community centers

Usage

getMembership(R, vertices, K, eig.values, eig.vectors)

Arguments

R	n-by-(K-1) ratio matrix
vertices	K-by-(K-1) community centers
K	number of communities.
eig.values	eigenvalues of adjacency matrix.
eig.vectors	eigenvectors of adjacency matrix.

Value

n-by-K membership matrix

---

Membership estimation algorithm called mixedSCORE

Description

Membership estimation algorithm called mixedSCORE

Usage

mixedSCORE(A, K, verbose = F)

Arguments

A	n-by-n binary symmtric adjacency matrix.
K	number of communities.
verbose	whether generate message

Value

A list containing

R

n-by-(K-1) ratio matrix.

L

Selected tunning parameter used for vertex hunting algorithm.

thetas

A vector of the estimated degree heterogeniety parameters

vertices

K-by-(K-1) K vertices of the found convex hull

centers

L-by-(K-1) L centers by kmeans

memberships

n-by-K membership matrix.

purity

A vector of maximum membership of each node

hard.cluster.labels

A vector of integers indicating hard clutering labels, by assigning the node to the cluster with max membership

Examples

library(igraphdata)
library(igraph)
data('karate')
A = get.adjacency(karate)
karate.mixed.out = mixedSCORE(A, 2)
karate.mixed.out$memberships

---

community detection method called SCORE Spectral Clustering On Ratios-of-Eigenvectors (SCORE)

Description

community detection method called SCORE Spectral Clustering On Ratios-of-Eigenvectors (SCORE)

Usage

SCORE(A, K, threshold = NULL)

Arguments

A	n-by-n binary symmtric adjacency matrix.
K	number of communities.
threshold	(optional) the threshold of ratio matrix. By defalt is log(n).

Value

A list containing

R

n-by-(K-1) ratio matrix.

labels

A vector of integer indicating the cluster to which each point allocated.

Examples

library(igraphdata)
library(igraph)
data('karate')
A = get.adjacency(karate)
karate.out = SCORE(A, 2)
karate.out$labels

---

community detection method called SCORE+

Description

community detection method called SCORE+

Usage

SCOREplus(A, k, c = 0.1, r = NULL)

Arguments

A	n-by-n binary symmtric adjacency matrix.
k	number of communities (>1).
c	(optional) tuning parameter for Graph Laplacian, default is 0.1.
r	(optional) latent dimension (>1), if not given, chosen between k and k+1 determined by eigen gap

Value

A list containing

label

Predicted community labels

ratios

n-by-(K-1) or n-by-r ratio matrix.

delta

calculated delta parameter

eig.vec

Top r eigen vectors

eig.val

Top r eigen values

Examples

library(igraphdata)
library(igraph)
data('karate')
A = get.adjacency(karate)
karate.plus.out = SCOREplus(A, 2)
karate.plus.out$labels

---

Vertex hunting algorithm to find the cluster centers

Description

Vertex hunting algorithm to find the cluster centers

Usage

vertexHunting(R, K, verbose = F)

Arguments

R	n-by-(K-1) ratio matrix
K	number of communities.
verbose	whether or not to show a progress bar

---

select the K vertices from given L centers

Description

select the K vertices from given L centers

Usage

vertexSearch(centers, K)

Arguments

centers	L-by-(K-1) center matrix
K	number of communities.

Value

A list containing

ind

a vector of K integers indicating the index of selected K vertices out of L centers.

dist

The maximum distance from centers to the convex hull formed by the K selected vertice

---