package com.etsy.conjecture.model;
import com.etsy.conjecture.data.LabeledInstance;
import com.etsy.conjecture.data.RealValuedLabel;
import com.etsy.conjecture.data.StringKeyedVector;
public class LeastSquaresRegressionModel extends
UpdateableLinearModel<RealValuedLabel> {
private static final long serialVersionUID = 1L;
public LeastSquaresRegressionModel(SGDOptimizer optimizer) {
super(optimizer);
}
public LeastSquaresRegressionModel(StringKeyedVector param, SGDOptimizer optimizer) {
super(param, optimizer);
}
@Override
public RealValuedLabel predict(StringKeyedVector instance) {
return new RealValuedLabel(param.dot(instance));
}
@Override
public double loss (LabeledInstance<RealValuedLabel> instance) {
double label = instance.getLabel().getValue();
double hypothesis = param.dot(instance.getVector());
return 0.5 * (hypothesis - label) * (hypothesis - label);
}
@Override
public StringKeyedVector getGradients(LabeledInstance<RealValuedLabel> instance) {
StringKeyedVector gradients = instance.getVector().copy();
double hypothesis = param.dot(instance.getVector());
double label = instance.getLabel().getValue();
gradients.mul((2 * (hypothesis-label)));
return gradients;
}
@Override
protected String getModelType() {
return "least_squares_regression";
}
}