com.expleague.ml.loss.FBetaLogit 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.ml.data.set.DataSet;

/**
 * User: solar
 * User: starlight
 * Date: 10.09.13
 * Time: 18:08
 */
public class FBetaLogit extends LLLogit {
  private final PLogit precision;
  private final RLogit recall;
  private final double betta;

  public FBetaLogit(final Vec target, final DataSet owner, final double betta) {
    super(target, owner);
    this.betta = betta;
    this.precision = new PLogit(target, owner);
    this.recall = new RLogit(target, owner);
  }

  /**
   * @param point $y=-log(\frac{1}{p(x|c)} - 1)$ supposed to be there so that $p(x|c) = \frac{1}{1+e^{-x}}$
   */
  @Override
  public double value(final Vec point) {
    final double precisionValue = precision.value(point);
    final double recallValue = recall.value(point);
    return (1 + betta * betta) * precisionValue * recallValue / (betta * betta * precisionValue + recallValue);
  }
}