com.expleague.ml.loss.WeightedLoss Maven / Gradle / Ivy

Go to download
Show more of this group Show more artifacts with this name
Show all versions of jmll
Various ML methods implemented by myself and my students
There is a newer version: 1.4.9
package com.expleague.ml.loss;

import com.expleague.commons.math.vectors.Vec;
import com.expleague.commons.func.AdditiveStatistics;
import com.expleague.commons.func.Factory;
import com.expleague.commons.math.Func;
import com.expleague.commons.math.Trans;
import com.expleague.ml.data.set.DataSet;
import gnu.trove.list.array.TIntArrayList;
import org.jetbrains.annotations.Nullable;

/**
 * User: solar
 * Date: 26.11.13
 * Time: 9:54
 */
public class WeightedLoss extends Func.Stub implements StatBasedLoss {
  private final BasedOn metric;
  private final int[] weights;

  public WeightedLoss(final BasedOn metric, final int[] weights) {
    this.metric = metric;
    this.weights = weights;
  }

  @Override
  public Factory statsFactory() {
    return () -> new Stat(weights, (AdditiveStatistics) metric.statsFactory().create());
  }

  @Override
  public Vec target() {
    return metric.target();
  }

  @Override
  public double bestIncrement(final Stat comb) {
    return metric.bestIncrement(comb.inside);
  }

  @Override
  public double score(final Stat comb) {
    return metric.score(comb.inside);
  }

  @Override
  public double value(final Stat comb) {
    return metric.value(comb.inside);
  }

  @Override
  public int dim() {
    return metric.xdim();
  }

  @Nullable
  @Override
  public Trans gradient() {
    return metric.gradient();
  }

  @Override
  public double value(final Vec x) {
    return metric.trans(x).get(0);
  }

  public double weight(final int index) {
    return weights[index];
  }

  public BasedOn base() {
    return metric;
  }

  @Override
  public DataSet owner() {
    return metric.owner();
  }

  public int[] points() {
    final TIntArrayList result = new TIntArrayList(weights.length + 1000); // Julian ????
    for(int i = 0; i < weights.length; i++) {
      if (weights[i] > 0)
        result.add(i);
    }
    return result.toArray();
  }

  public int[] zeroPoints() {
    final TIntArrayList result = new TIntArrayList(weights.length);
    for(int i = 0; i < weights.length; i++) {
      if (weights[i] == 0)
        result.add(i);
    }
    return result.toArray();
  }

  public static class Stat implements AdditiveStatistics {
    public AdditiveStatistics inside;
    private final int[] weights;

    public Stat(final int[] weights, final AdditiveStatistics inside) {
      this.weights = weights;
      this.inside = inside;
    }

    @Override
    public Stat append(final int index, final int times) {
      final int count = weights[index];
      inside.append(index, count * times);
      return this;
    }

    @Override
    public Stat append(final AdditiveStatistics other) {
      inside.append(((Stat) other).inside);
      return this;
    }

    @Override
    public Stat remove(final int index, final int times) {
      final int count = weights[index];
      inside.remove(index, count * times);
      return this;
    }

    @Override
    public Stat remove(final AdditiveStatistics other) {
      inside.remove(((Stat) other).inside);
      return this;
    }

    @Override
    public Stat append(int index, double weight) {
      final int count = weights[index];
      inside.append(index,weight*count);
      return this;
    }

    @Override
    public Stat remove(int index, double weight) {
      final int count = weights[index];
      inside.remove(index,weight*count);
      return this;
    }
  }
}