Class SDCA

JSAT/src/jsat/classifiers/linear/SDCA.java:64–644

This class implements the Proximal Stochastic Dual Coordinate Ascent (SDCA) algorithm for learning general linear models with Elastic-Net regularization. It is a fast learning algorithm, and can be used for generic Logistic or least-squares regression with elastic-net regularization. It can work

 * @author Edward Raff <Raff.Edward@gmail.com>
 */
public class SDCA implements Classifier, Regressor, Parameterized, SimpleWeightVectorModel, WarmClassifier, WarmRegressor
{
    private LossFunc loss;
    private boolean useBias = true;
    private double tol = 0.001;
    private double lambda;
    private double alpha = 0.5;
    private int max_epochs = 200;
    private double[] dual_alphas;
    /**
     * Returns the number of epochs SDCA took until reaching convergence. 
     */
    protected int epochs_taken;
    
    private Vec[] ws;
    private double[] bs;

    /**
     * Creates a new SDCA learner for {@link LogisticLoss logistic-regression}. 
     * Pure L2 or L1 regularization can be obtained using the
     * {@link #setAlpha(double) alpha} parameter.
     */
    public SDCA()
    {
        this(1e-5);
    }

    /**
     * <br>The implementation will use Elastic-Net regularization by default.
     * Pure L2 or L1 regularization can be obtained using the
     * {@link #setAlpha(double) alpha} parameter.
     * @param lambda the regularization penalty to use. 
     */
    public SDCA(double lambda)
    {
        this(lambda, new LogisticLoss());
    }

    /**
     * Creates a new SDCA learner for either a classification or regression
     * problem, the type of which is determined by the LossFunction given.
     * <br>The implementation will use Elastic-Net regularization by default.
     * Pure L2 or L1 regularization can be obtained using the
     * {@link #setAlpha(double) alpha} parameter.
     *
     * @param lambda the regularization penalty to use.
     * @param loss the loss function to use for training, which determines if it
     * implements classification or regression
     */
    public SDCA(double lambda, LossFunc loss)
    {
        setLoss(loss);
        setLambda(lambda);
    }
    

    /**
     * Copy Constructor