Package 'conclust'

Pairwise Constraints Clustering

Description

There are 4 main functions in this package: ckmeans(), lcvqe(), mpckm() and ccls(). They take an unlabeled dataset and two lists of must-link and cannot-link constraints as input and produce a clustering as output.

Details

The DESCRIPTION file:

Package:	conclust
Type:	Package
Title:	Pairwise Constraints Clustering
Version:	1.1
Date:	2016-08-15
Author:	Tran Khanh Hiep, Nguyen Minh Duc
Maintainer:	Tran Khanh Hiep <[email protected]>
Description:	There are 4 main functions in this package: ckmeans(), lcvqe(), mpckm() and ccls(). They take an unlabeled dataset and two lists of must-link and cannot-link constraints as input and produce a clustering as output.
License:	GPL-3
NeedsCompilation:	no
Packaged:	2016-08-15 08:45:24 UTC; Hiep
Repository:	CRAN
Date/Publication:	2016-08-15 13:16:21

Index of help topics:

ccls                    Pairwise Constrained Clustering by Local Search
ckmeans                 COP K-means algorithm
conclust-package        Pairwise Constraints Clustering
lcvqe                   LCVQE algorithm
mpckm                   MPC K-means algorithm

There are 4 main functions in this package: ckmeans(), lcvqe(), mpckm() and ccls(). They take an unlabeled dataset and two lists of must-link and cannot-link constraints as input and produce a clustering as output.

Author(s)

Tran Khanh Hiep, Nguyen Minh Duc

Maintainer: Tran Khanh Hiep <[email protected]>

References

Wagstaff, Cardie, Rogers, Schrodl (2001), Constrained K-means Clustering with Background Knowledge Bilenko, Basu, Mooney (2004), Integrating Constraints and Metric Learning in Semi-Supervised Clustering Dan Pelleg, Dorit Baras (2007), K-means with large and noisy constraint sets

See Also

Wagstaff, Cardie, Rogers, Schrodl (2001), Constrained K-means Clustering with Background Knowledge Bilenko, Basu, Mooney (2004), Integrating Constraints and Metric Learning in Semi-Supervised Clustering Dan Pelleg, Dorit Baras (2007), K-means with large and noisy constraint sets

Examples

data = matrix(c(0, 1, 1, 0, 0, 0, 1, 1), nrow = 4)
mustLink = matrix(c(1, 2), nrow = 1)
cantLink = matrix(c(1, 4), nrow = 1)
k = 2
pred = ckmeans(data, k, mustLink, cantLink)
pred
pred = mpckm(data, k, mustLink, cantLink)
pred
pred = lcvqe(data, k, mustLink, cantLink)
pred
pred = ccls(data, k, mustLink, cantLink)
pred

---

Pairwise Constrained Clustering by Local Search

Description

This function takes an unlabeled dataset and two lists of must-link and cannot-link constraints as input and produce a clustering as output.

Usage

ccls(data, k = -1, mustLink, cantLink, maxIter = 1, tabuIter = 100, tabuLength = 20)

Arguments

data	The unlabeled dataset.
k	Number of clusters.
mustLink	A list of must-link constraints
cantLink	A list of cannot-link constraints
maxIter	Number of iteration
tabuIter	Number of iteration in Tabu search
tabuLength	The number of elements in the Tabu list

Details

This algorithm minimizes the clustering cost function using Tabu search.

Value

A vector that represents the labels (clusters) of the data points

Note

This is the first algorithm for pairwise constrained clustering by local search.

Author(s)

Tran Khanh Hiep Nguyen Minh Duc

References

Tran Khanh Hiep, Nguyen Minh Duc, Bui Quoc Trung (2016), Pairwise Constrained Clustering by Local Search.

See Also

Tran Khanh Hiep, Nguyen Minh Duc, Bui Quoc Trung (2016), Pairwise Constrained Clustering by Local Search.

Examples

data = matrix(c(0, 1, 1, 0, 0, 0, 1, 1), nrow = 4)
mustLink = matrix(c(1, 2), nrow = 1)
cantLink = matrix(c(1, 4), nrow = 1)
k = 2
pred = ckmeans(data, k, mustLink, cantLink)
pred

---

COP K-means algorithm

Description

This function takes an unlabeled dataset and two lists of must-link and cannot-link constraints as input and produce a clustering as output.

Usage

ckmeans(data, k, mustLink, cantLink, maxIter = 100)

Arguments

data	The unlabeled dataset.
k	Number of clusters.
mustLink	A list of must-link constraints
cantLink	A list of cannot-link constraints
maxIter	Number of iteration

Details

This algorithm produces a clustering that satisfies all given constraints.

Value

A vector that represents the labels (clusters) of the data points

Note

The constraints should be consistent in order for the algorithm to work.

Author(s)

Tran Khanh Hiep Nguyen Minh Duc

References

Wagstaff, Cardie, Rogers, Schrodl (2001), Constrained K-means Clustering with Background Knowledge

See Also

Wagstaff, Cardie, Rogers, Schrodl (2001), Constrained K-means Clustering with Background Knowledge

Examples

data = matrix(c(0, 1, 1, 0, 0, 0, 1, 1), nrow = 4)
mustLink = matrix(c(1, 2), nrow = 1)
cantLink = matrix(c(1, 4), nrow = 1)
k = 2
pred = ckmeans(data, k, mustLink, cantLink)
pred

---

LCVQE algorithm

Description

This function takes an unlabeled dataset and two lists of must-link and cannot-link constraints as input and produce a clustering as output.

Usage

lcvqe(data, k, mustLink, cantLink, maxIter = 10)

Arguments

data	The unlabeled dataset.
k	Number of clusters.
mustLink	A list of must-link constraints
cantLink	A list of cannot-link constraints
maxIter	Number of iteration

Details

This algorithm finds a clustering that satisfies as many constraints as possible

Value

A vector that represents the labels (clusters) of the data points

Note

This algorithm can handle noisy constraints.

Author(s)

Tran Khanh Hiep Nguyen Minh Duc

References

Dan Pelleg, Dorit Baras (2007), K-means with large and noisy constraint sets

See Also

Dan Pelleg, Dorit Baras (2007), K-means with large and noisy constraint sets

Examples

data = matrix(c(0, 1, 1, 0, 0, 0, 1, 1), nrow = 4)
mustLink = matrix(c(1, 2), nrow = 1)
cantLink = matrix(c(1, 4), nrow = 1)
k = 2
pred = lcvqe(data, k, mustLink, cantLink)
pred

---

MPC K-means algorithm

Description

This function takes an unlabeled dataset and two lists of must-link and cannot-link constraints as input and produce a clustering as output.

Usage

mpckm(data, k, mustLink, cantLink, maxIter = 10)

Arguments

data	The unlabeled dataset.
k	Number of clusters.
mustLink	A list of must-link constraints
cantLink	A list of cannot-link constraints
maxIter	Number of iteration

Details

This algorithm finds a clustering that satisfies as many constraints as possible

Value

A vector that represents the labels (clusters) of the data points

Note

This is one of the best algorithm for clustering with constraints.

Author(s)

Tran Khanh Hiep Nguyen Minh Duc

References

Bilenko, Basu, Mooney (2004), Integrating Constraints and Metric Learning in Semi-Supervised Clustering

See Also

Bilenko, Basu, Mooney (2004), Integrating Constraints and Metric Learning in Semi-Supervised Clustering

Examples

data = matrix(c(0, 1, 1, 0, 0, 0, 1, 1), nrow = 4)
mustLink = matrix(c(1, 2), nrow = 1)
cantLink = matrix(c(1, 4), nrow = 1)
k = 2
pred = mpckm(data, k, mustLink, cantLink)
pred

---