edu.ucla.sspace.clustering.criterion.E1Function Maven / Gradle / Ivy

Go to download

Show more of this group Show more artifacts with this name
Show all versions of sspace-wordsi Show documentation

The S-Space Package is a collection of algorithms for building Semantic Spaces as well as a highly-scalable library for designing new distributional semantics algorithms. Distributional algorithms process text corpora and represent the semantic for words as high dimensional feature vectors. This package also includes matrices, vectors, and numerous clustering algorithms. These approaches are known by many names, such as word spaces, semantic spaces, or distributed semantics and rest upon the Distributional Hypothesis: words that appear in similar contexts have similar meanings.

The newest version!

/*
 * Copyright 2011 Keith Stevens 
 *
 * This file is part of the S-Space package and is covered under the terms and
 * conditions therein.
 *
 * The S-Space package is free software: you can redistribute it and/or modify
 * it under the terms of the GNU General Public License version 2 as published
 * by the Free Software Foundation and distributed hereunder to you.
 *
 * THIS SOFTWARE IS PROVIDED "AS IS" AND NO REPRESENTATIONS OR WARRANTIES,
 * EXPRESS OR IMPLIED ARE MADE.  BY WAY OF EXAMPLE, BUT NOT LIMITATION, WE MAKE
 * NO REPRESENTATIONS OR WARRANTIES OF MERCHANT- ABILITY OR FITNESS FOR ANY
 * PARTICULAR PURPOSE OR THAT THE USE OF THE LICENSED SOFTWARE OR DOCUMENTATION
 * WILL NOT INFRINGE ANY THIRD PARTY PATENTS, COPYRIGHTS, TRADEMARKS OR OTHER
 * RIGHTS.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program. If not, see .
 */

package edu.ucla.sspace.clustering.criterion;

import edu.ucla.sspace.common.Similarity;

import edu.ucla.sspace.matrix.Matrix;

import edu.ucla.sspace.vector.DenseDynamicMagnitudeVector;
import edu.ucla.sspace.vector.DoubleVector;
import edu.ucla.sspace.vector.VectorMath;

import java.util.List;


/**
 * This {@link CriterionFunction} measures the external differences between
 * clusters.  It gives a better score to clustering solutions with centroids
 * that are more distant from the centroid for the data set as a whole.
 *
 * @author Keith Stevens
 */
public class E1Function extends BaseFunction {

    /**
     * The centroid for all data points if they were assigned to a single
     * cluster.
     */
    private DoubleVector completeCentroid;

    /**
     * The the dot product between each cluster and {@code completeCentroid}.
     */
    private double[] simToComplete;

    /**
     * Constructs a new {@link E1Function}.
     */
    public E1Function() {
    }

    /**
     * A package private constructor for all {@link CriterionFunction}s
     * subclassing from this {@link BaseFunction}.  This is to facilitate the
     * implementation of {@link HybridBaseFunction}.  The provided objects are
     * intended to replace those that would have been computed by {@link
     * #setup(Matrix, int[], int) setup} so that one class can do this work once
     * and then share the computed values with other functions.
     *
     * @param matrix The list of normalized data points that are to be
     *        clustered
     * @param centroids The set of centroids associated with the dataset.
     * @param costs The set of costs for each centroid.
     * @param assignments The initial assignments for each cluster.
     * @param clusterSizes The size of each cluster.
     * @param completeCentroid The new summation vector of all data points
     * @param simToComplete The distance from each cluster to {@code
     *        completeCentroid}
     */
    E1Function(List matrix,
               DoubleVector[] centroids,
               double[] costs,
               int[] assignments,
               int[] clusterSizes,
               DoubleVector completeCentroid,
               double[] simToComplete) {
        super(matrix, centroids, costs, assignments, clusterSizes);
        this.completeCentroid = completeCentroid;
        this.simToComplete = simToComplete;
    }

    /**
     * {@inheritDoc}
     */
    protected void subSetup(Matrix m) {
        completeCentroid = new DenseDynamicMagnitudeVector(m.rows());
        for (DoubleVector v : matrix)
            VectorMath.add(completeCentroid, v);

        simToComplete = new double[centroids.length];
        for (int c = 0; c < centroids.length; ++c)
            simToComplete[c] = VectorMath.dotProduct(
                    centroids[c], completeCentroid);
    }

    /**
     * {@inheritDoc}
     */
    protected double getOldCentroidScore(DoubleVector vector,
                                         int oldCentroidIndex,
                                         int altClusterSize) {
        double newScore = simToComplete[oldCentroidIndex];
        newScore -= VectorMath.dotProduct(completeCentroid, vector);
        newScore /= subtractedMagnitude(centroids[oldCentroidIndex], vector);
        newScore *= altClusterSize;
        return newScore;
    }

    /**
     * {@inheritDoc}
     */
    protected double getNewCentroidScore(int newCentroidIndex,
                                         DoubleVector dataPoint) {
        double newScore = VectorMath.dotProduct(completeCentroid, dataPoint);
        newScore += simToComplete[newCentroidIndex];
        newScore /= modifiedMagnitude(centroids[newCentroidIndex], dataPoint);
        newScore *= (clusterSizes[newCentroidIndex] + 1);
        return newScore;
    }

    /**
     * {@inheritDoc}
     */
    public boolean isMaximize() {
        return false;
    }

    /**
     * {@inheritDoc}
     */
    protected void updateScores(int newCentroidIndex,
                                int oldCentroidIndex,
                                DoubleVector vector) {
        simToComplete[newCentroidIndex] += VectorMath.dotProduct(
                completeCentroid, vector);
        simToComplete[oldCentroidIndex] -= VectorMath.dotProduct(
                completeCentroid, vector);
    }
}