com.expleague.ml.loss.ExclusiveComplementLLLogit Maven / Gradle / Ivy
package com.expleague.ml.loss;
import com.expleague.commons.math.vectors.Vec;
import com.expleague.commons.math.vectors.impl.vectors.ArrayVec;
import com.expleague.commons.math.FuncC1;
import com.expleague.ml.TargetFunc;
import com.expleague.ml.data.set.DataSet;
import static java.lang.Math.exp;
import static java.lang.Math.log;
/**
* User: solar
* Date: 19.03.15
* Time: 13:43
*/
public class ExclusiveComplementLLLogit extends FuncC1.Stub implements TargetFunc {
private final double alpha;
protected final Vec target;
private final Vec complimentProbab;
private final DataSet> owner;
public ExclusiveComplementLLLogit(double alpha, final Vec target, final Vec complimentProbab, final DataSet> owner) {
this.alpha = alpha;
this.target = target;
this.complimentProbab = complimentProbab;
this.owner = owner;
}
@Override
public Vec gradient(final Vec x) {
final Vec result = new ArrayVec(x.dim());
final double a = alpha;
for (int i = 0; i < x.dim(); i++) {
final double expX = exp(x.get(i));
final double pX = expX / (1 + expX);
double c = complimentProbab.get(i) ;
if (target.get(i) > 0) // positive example
result.set(i, (a - 1) * pX / (a + expX));
else
result.set(i, c * (a - 1) * pX / (a * c + (c - 1) * expX - 1));
}
return result;
}
@Override
public int dim() {
return target.dim();
}
@Override
public double value(final Vec point) {
double result = 0;
for (int i = 0; i < point.dim(); i++) {
final double expMX = exp(-point.get(i));
final double pX = 1. / (1. + expMX);
final double complimentP = complimentProbab.get(i);
final double pGen = complimentP * (alpha + (1 - alpha) * pX);
if (target.get(i) > 0) // positive example
result += log(pGen);
else // negative
result += log(1 - pGen);
}
return exp(result / point.dim());
}
public int label(final int idx) {
return (int)target.get(idx);
}
public Vec labels() {
return target;
}
@Override
public DataSet> owner() {
return owner;
}
}