Class DANN

JSAT/src/jsat/classifiers/knn/DANN.java:42–462

DANN is an implementation of Discriminant Adaptive Nearest Neighbor. DANN has a fixed O(n) classification time. At each classification, DANN uses a large set of points to iteratively create and adjust a distance metic that reflects the separability of classes at a localized level. This increa

 * @author Edward Raff
 */
public class DANN implements Classifier, Parameterized
{

    private static final long serialVersionUID = -272865942127664672L;
    /**
     * The default number of neighbors to use when building a metric is 
     * {@value #DEFAULT_KN}. 
     */
    public static final int DEFAULT_KN = 40;
    /**
     * The default number of neighbors to use when classifying is 
     * {@value #DEFAULT_K}
     */
    public static final int DEFAULT_K = 1;
    /**
     * The default regularization used when building a metric is 
     * {@value #DEFAULT_EPS}
     */
    public static final double DEFAULT_EPS = 1.0;
    /**
     * The default number of iterations for creating the metric is
     * {@value #DEFAULT_ITERATIONS}
     */
    public static final int DEFAULT_ITERATIONS = 1;
    
    private int kn;
    private int k;
    private int maxIterations;
    private double eps;
    
    private VectorCollectionFactory<VecPaired<Vec, Integer>> vcf;
    
    private CategoricalData predicting;
    
    /**
     * Vectors paired with their index in the original data set
     */
    private VectorCollection<VecPaired<Vec, Integer>> vc;
    private List<VecPaired<Vec, Integer>> vecList;
    
    /**
     * Creates a new DANN classifier
     */
    public DANN()
    {
        this(DEFAULT_KN, DEFAULT_K);
    }
    
    /**
     * Creates a new DANN classifier
     * @param kn the number of neighbors to use in casting a net to build a new metric
     * @param k the number of neighbors to use with the final metric in classification
     */
    public DANN(int kn, int k)
    {
        this(kn, k, DEFAULT_EPS);
    }