smile.math.distance

Class MinkowskiDistance

java.lang.Object

smile.math.distance.MinkowskiDistance

All Implemented Interfaces:

Distance<double[]>, Metric<double[]>

public class MinkowskiDistance
extends Object
implements Metric<double[]>

Minkowski distance of order p or L_p-norm, is a generalization of Euclidean distance that is actually L₂-norm. You may also provide a specified weight vector. For float or double arrays, missing values (i.e. NaN) are also handled. Also support sparse arrays of which zeros are excluded to save space.

Author:

Haifeng Li

Constructor Summary

Constructors

Constructor and Description
MinkowskiDistance(int p) Constructor.
MinkowskiDistance(int p, double[] weight) Constructor.

Method Summary

Methods

Modifier and Type	Method and Description
double	d(double[] x, double[] y) Minkowski distance between the two arrays of type double.
double	d(float[] x, float[] y) Minkowski distance between the two arrays of type float.
double	d(int[] x, int[] y) Minkowski distance between the two arrays of type integer.
String	toString()

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, wait, wait, wait

Constructor Detail

MinkowskiDistance

public MinkowskiDistance(int p)

Constructor.

MinkowskiDistance

public MinkowskiDistance(int p,
                 double[] weight)

Constructor.

Parameters:

weight - the weight vector.

Method Detail

toString

public String toString()

Overrides:

toString in class Object

d

public double d(int[] x,
       int[] y)

Minkowski distance between the two arrays of type integer.

d

public double d(float[] x,
       float[] y)

Minkowski distance between the two arrays of type float. NaN will be treated as missing values and will be excluded from the calculation. Let m be the number non-missing values, and n be the number of all values. The returned distance is pow(n * d / m, 1/p), where d is the p-pow of distance between non-missing values.

d

public double d(double[] x,
       double[] y)

Minkowski distance between the two arrays of type double. NaN will be treated as missing values and will be excluded from the calculation. Let m be the number non-missing values, and n be the number of all values. The returned distance is pow(n * d / m, 1/p), where d is the p-pow of distance between non-missing values.

Specified by:

d in interface Distance<double[]>