edu.stanford.nlp.cluster
Class SimilarityMatrix

java.lang.Object
  |
  +--edu.stanford.nlp.cluster.SimilarityMatrix

All Implemented Interfaces:

MatrixWrapper

public class SimilarityMatrix

extends Object

implements MatrixWrapper

Variables and methods for constructing a matrix of similarities of a data collection. The client must provide the Similarity metric.

Field Summary

protected ArrayListMatrix

similarities

Constructor Summary

SimilarityMatrix(Clusters c)
Creates new SimilarityMatrix with default Cosine Similarity

SimilarityMatrix(Matrix m)
Creates new SimilarityMatrix with default Cosine Similarity

SimilarityMatrix(Similarity s, Clusters c)
Creates new SimilarityMatrix with Similarity Metric defined by client

SimilarityMatrix(Similarity s, Matrix m)
Creates new SimilarityMatrix with Similarity Metric defined by client

Method Summary

Matrix	createSimilarityMatrix(Similarity s, Clusters c) create a full dense matrix of similarities, where similarities(i,j) gives the similarity between the probability distribution over data of clusters i and j of Clusters c as determined by the similarity metric s
Matrix	createSimilarityMatrix(Similarity s, Matrix m) create a full dense matrix of similarities, where similarities(i,j) gives the similarity between columns i and j of Matrix m as determined by the similarity metric s
Matrix	getMatrix() Returns underlying Matrix
void	initialize(int size) initializes similarities matrix to be a m x m square matrix, where m=size the matrix is of type ArrayListMatrix, the columns of which are of type GrowableDenseDoubleArray
void	merge(int i, int j, Double4Function csf, Clusters clusters) Merges columns i,j in the similarity matrix by setting the value at each entry k in the column to be csf.apply(i(k),j(k)).
void	merge(int i, int j, cern.colt.function.DoubleDoubleFunction csf) Merges columns i,j in the similarity matrix by setting the value at each entry k in the column to be csf.apply(i(k),j(k)).

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Detail

similarities

protected ArrayListMatrix similarities

Constructor Detail

SimilarityMatrix

public SimilarityMatrix(Similarity s,
                        Matrix m)

Creates new SimilarityMatrix with Similarity Metric defined by client

SimilarityMatrix

public SimilarityMatrix(Similarity s,
                        Clusters c)

Creates new SimilarityMatrix with Similarity Metric defined by client

SimilarityMatrix

public SimilarityMatrix(Matrix m)

Creates new SimilarityMatrix with default Cosine Similarity

SimilarityMatrix

public SimilarityMatrix(Clusters c)

Creates new SimilarityMatrix with default Cosine Similarity

Method Detail

initialize

public void initialize(int size)

initializes similarities matrix to be a m x m square matrix, where m=size the matrix is of type ArrayListMatrix, the columns of which are of type GrowableDenseDoubleArray

createSimilarityMatrix

public Matrix createSimilarityMatrix(Similarity s,
                                     Matrix m)

create a full dense matrix of similarities, where similarities(i,j) gives the similarity between columns i and j of Matrix m as determined by the similarity metric s

createSimilarityMatrix

public Matrix createSimilarityMatrix(Similarity s,
                                     Clusters c)

create a full dense matrix of similarities, where similarities(i,j) gives the similarity between the probability distribution over data of clusters i and j of Clusters c as determined by the similarity metric s

getMatrix

public Matrix getMatrix()

Description copied from interface: MatrixWrapper

Returns underlying Matrix

Specified by:

getMatrix in interface MatrixWrapper

merge

public void merge(int i,
                  int j,
                  cern.colt.function.DoubleDoubleFunction csf)

Merges columns i,j in the similarity matrix by setting the value at each entry k in the column to be csf.apply(i(k),j(k)). This corresponds to single-link or complete-link agglomerative clustering. Merged column takes the spot of the column with the lower index, and the column (and row) with the higher index is removed. DoubleDoubleFunction csf is the Cluster Similarity Function (Single-Link, Complete-Link, etc.)

merge

public void merge(int i,
                  int j,
                  Double4Function csf,
                  Clusters clusters)

Merges columns i,j in the similarity matrix by setting the value at each entry k in the column to be csf.apply(i(k),j(k)). This corresponds to group-average (or related) agglomerative clustering Merged column takes the spot of the column with the lower index, and the column (and row) with the higher index is removed. Double4Function csf is the Cluster Similarity Function (Group-Average, etc.)