path: root/src/de/lmu/ifi/dbs/elki/datasource/filter/normalization/AttributeWiseCDFNormalization.java

package de.lmu.ifi.dbs.elki.datasource.filter.normalization;

/*
 This file is part of ELKI:
 Environment for Developing KDD-Applications Supported by Index-Structures

 Copyright (C) 2013
 Ludwig-Maximilians-Universität München
 Lehr- und Forschungseinheit für Datenbanksysteme
 ELKI Development Team

 This program is free software: you can redistribute it and/or modify
 it under the terms of the GNU Affero General Public License as published by
 the Free Software Foundation, either version 3 of the License, or
 (at your option) any later version.

 This program is distributed in the hope that it will be useful,
 but WITHOUT ANY WARRANTY; without even the implied warranty of
 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 GNU Affero General Public License for more details.

 You should have received a copy of the GNU Affero General Public License
 along with this program.  If not, see <http://www.gnu.org/licenses/>.
 */

import java.util.ArrayList;
import java.util.Arrays;
import java.util.List;

import de.lmu.ifi.dbs.elki.data.NumberVector;
import de.lmu.ifi.dbs.elki.data.type.SimpleTypeInformation;
import de.lmu.ifi.dbs.elki.data.type.TypeUtil;
import de.lmu.ifi.dbs.elki.data.type.VectorFieldTypeInformation;
import de.lmu.ifi.dbs.elki.datasource.bundle.MultipleObjectsBundle;
import de.lmu.ifi.dbs.elki.datasource.filter.FilterUtil;
import de.lmu.ifi.dbs.elki.logging.Logging;
import de.lmu.ifi.dbs.elki.math.linearalgebra.LinearEquationSystem;
import de.lmu.ifi.dbs.elki.math.statistics.distribution.Distribution;
import de.lmu.ifi.dbs.elki.math.statistics.distribution.estimator.DistributionEstimator;
import de.lmu.ifi.dbs.elki.math.statistics.distribution.estimator.meta.BestFitEstimator;
import de.lmu.ifi.dbs.elki.math.statistics.tests.KolmogorovSmirnovTest;
import de.lmu.ifi.dbs.elki.utilities.datastructures.arraylike.NumberArrayAdapter;
import de.lmu.ifi.dbs.elki.utilities.exceptions.ExceptionMessages;
import de.lmu.ifi.dbs.elki.utilities.optionhandling.AbstractParameterizer;
import de.lmu.ifi.dbs.elki.utilities.optionhandling.OptionID;
import de.lmu.ifi.dbs.elki.utilities.optionhandling.parameterization.Parameterization;
import de.lmu.ifi.dbs.elki.utilities.optionhandling.parameters.ObjectListParameter;

/**
 * Class to perform and undo a normalization on real vectors by estimating the
 * distribution of values along each dimension independently, then rescaling
 * objects to the cumulative density function (CDF) value at the original
 * coordinate.
 * 
 * This process is for example also discussed in section 3.4 of
 * <p>
 * Effects of Feature Normalization on Image Retrieval <br/>
 * S. Aksoy, R. M. Haralick
 * </p>
 * but they do not detail how to obtain an appropriate function `F`.
 * 
 * @author Erich Schubert
 * @param <V> vector type
 * 
 * @apiviz.uses NumberVector
 * @apiviz.uses DistributionEstimator
 */
// TODO: extract superclass AbstractAttributeWiseNormalization
public class AttributeWiseCDFNormalization<V extends NumberVector<?>> implements Normalization<V> {
  /**
   * Class logger.
   */
  private static final Logging LOG = Logging.getLogger(AttributeWiseCDFNormalization.class);

  /**
   * Stores the distribution estimators
   */
  private List<DistributionEstimator<?>> estimators;

  /**
   * Stores the estimated distributions
   */
  private List<Distribution> dists;

  /**
   * Number vector factory.
   */
  protected NumberVector.Factory<V, ?> factory;

  /**
   * Constructor.
   * 
   * @param estimators Distribution estimators
   */
  public AttributeWiseCDFNormalization(List<DistributionEstimator<?>> estimators) {
    super();
    this.estimators = estimators;
  }

  @Override
  public MultipleObjectsBundle filter(MultipleObjectsBundle objects) {
    if (objects.dataLength() == 0) {
      return objects;
    }
    for (int r = 0; r < objects.metaLength(); r++) {
      SimpleTypeInformation<?> type = (SimpleTypeInformation<?>) objects.meta(r);
      final List<?> column = (List<?>) objects.getColumn(r);
      if (!TypeUtil.NUMBER_VECTOR_FIELD.isAssignableFromType(type)) {
        continue;
      }
      @SuppressWarnings("unchecked")
      final List<V> castColumn = (List<V>) column;
      // Get the replacement type information
      @SuppressWarnings("unchecked")
      final VectorFieldTypeInformation<V> castType = (VectorFieldTypeInformation<V>) type;
      factory = FilterUtil.guessFactory(castType);

      // Scan to find the best
      final int dim = castType.getDimensionality();
      dists = new ArrayList<>(dim);
      // Scratch space for testing:
      double[] test = new double[castColumn.size()];

      // We iterate over dimensions, this kind of filter needs fast random
      // access.
      Adapter adapter = new Adapter();
      for (int d = 0; d < dim; d++) {
        adapter.dim = d;
        if (estimators.size() == 1) {
          dists.add(estimators.get(0).estimate(castColumn, adapter));
        } else {
          Distribution best = null;
          double bestq = Double.POSITIVE_INFINITY;
          trials: for (DistributionEstimator<?> est : estimators) {
            try {
              Distribution dist = est.estimate(castColumn, adapter);
              for (int i = 0; i < test.length; i++) {
                test[i] = dist.cdf(castColumn.get(i).doubleValue(d));
                if (Double.isNaN(test[i])) {
                  LOG.warning("Got NaN after fitting " + est.toString() + ": " + dist.toString());
                  continue trials;
                }
                if (Double.isInfinite(test[i])) {
                  LOG.warning("Got infinite value after fitting " + est.toString() + ": " + dist.toString());
                  continue trials;
                }
              }
              Arrays.sort(test);
              double q = KolmogorovSmirnovTest.simpleTest(test);
              if (LOG.isVeryVerbose()) {
                LOG.veryverbose("Estimator " + est.toString() + " (" + dist.toString() + ") has maximum deviation " + q + " for dimension " + d);
              }
              if (best == null || q < bestq) {
                best = dist;
                bestq = q;
              }
            } catch (ArithmeticException e) {
              if (LOG.isVeryVerbose()) {
                LOG.veryverbose("Fitting distribution " + est + " failed: " + e.getMessage());
              }
              continue;
            }
          }
          if (LOG.isVerbose()) {
            LOG.verbose("Best fit for dimension " + d + ": " + best.toString());
          }
          dists.add(best);
        }
      }

      // Normalization scan
      double[] buf = new double[dim];
      for (int i = 0; i < objects.dataLength(); i++) {
        final V obj = castColumn.get(i);
        for (int d = 0; d < dim; d++) {
          buf[d] = dists.get(d).cdf(obj.doubleValue(d));
        }
        castColumn.set(i, factory.newNumberVector(buf));
      }
    }
    return objects;
  }

  @Override
  public V restore(V featureVector) throws NonNumericFeaturesException {
    throw new UnsupportedOperationException(ExceptionMessages.UNSUPPORTED_NOT_YET);
  }

  @Override
  public LinearEquationSystem transform(LinearEquationSystem linearEquationSystem) {
    throw new UnsupportedOperationException(ExceptionMessages.UNSUPPORTED_NOT_YET);
  }

  @Override
  public String toString() {
    StringBuilder result = new StringBuilder();
    result.append("normalization class: ").append(getClass().getName());
    result.append('\n');
    result.append("normalization distributions: ");
    boolean first = true;
    for (DistributionEstimator<?> est : estimators) {
      if (!first) {
        result.append(',');
      }
      first = false;
      result.append(est.getClass().getSimpleName());
    }
    return result.toString();
  }

  /**
   * Array adapter class for vectors.
   * 
   * @author Erich Schubert
   *
   * @apiviz.exclude
   */
  private static class Adapter implements NumberArrayAdapter<Double, List<? extends NumberVector<?>>> {
    /**
     * Dimension to process.
     */
    int dim;

    @Override
    public int size(List<? extends NumberVector<?>> array) {
      return array.size();
    }

    @Override
    public Double get(List<? extends NumberVector<?>> array, int off) throws IndexOutOfBoundsException {
      return getDouble(array, off);
    }

    @Override
    public double getDouble(List<? extends NumberVector<?>> array, int off) throws IndexOutOfBoundsException {
      return array.get(off).doubleValue(dim);
    }

    @Override
    public float getFloat(List<? extends NumberVector<?>> array, int off) throws IndexOutOfBoundsException {
      return array.get(off).floatValue(dim);
    }

    @Override
    public int getInteger(List<? extends NumberVector<?>> array, int off) throws IndexOutOfBoundsException {
      return array.get(off).intValue(dim);
    }

    @Override
    public short getShort(List<? extends NumberVector<?>> array, int off) throws IndexOutOfBoundsException {
      return array.get(off).shortValue(dim);
    }

    @Override
    public long getLong(List<? extends NumberVector<?>> array, int off) throws IndexOutOfBoundsException {
      return array.get(off).longValue(dim);
    }

    @Override
    public byte getByte(List<? extends NumberVector<?>> array, int off) throws IndexOutOfBoundsException {
      return array.get(off).byteValue(dim);
    }
  }

  /**
   * Parameterization class.
   * 
   * @author Erich Schubert
   * 
   * @apiviz.exclude
   */
  public static class Parameterizer<V extends NumberVector<?>> extends AbstractParameterizer {
    /**
     * Parameter for distribution estimators.
     */
    public static final OptionID DISTRIBUTIONS_ID = new OptionID("normalize.distributions", "A list of the distribution estimators to try.");

    /**
     * Stores the distribution estimators
     */
    private List<DistributionEstimator<?>> estimators;

    @Override
    protected void makeOptions(Parameterization config) {
      super.makeOptions(config);
      ObjectListParameter<DistributionEstimator<?>> estP = new ObjectListParameter<>(DISTRIBUTIONS_ID, DistributionEstimator.class);
      List<Class<? extends DistributionEstimator<?>>> def = new ArrayList<>(1);
      def.add(BestFitEstimator.class);
      estP.setDefaultValue(def);
      if (config.grab(estP)) {
        estimators = estP.instantiateClasses(config);
      }
    }

    @Override
    protected AttributeWiseCDFNormalization<V> makeInstance() {
      return new AttributeWiseCDFNormalization<>(estimators);
    }
  }
}