程序包 org.meteoinfo.math.stats.kde

类 BallTree

java.lang.Object

org.meteoinfo.math.stats.kde.BallTree

public class BallTree
extends Object

字段概要

字段

限定符和类型	字段	说明
static int	numSkipped

构造器概要

构造器

限定符	构造器	说明
protected	BallTree(List<Event> data, int minParent)	Creates a Ball Tree (k-d tree) for the data to reduce the speed of the log probability computation.

方法概要

限定符和类型	方法	说明
protected static double	computeLogKernel(double[] y, double[] s, double h)
protected static double	computeLogKernel(double[] y, double[] s, double hMax, double hMin)	The actual math for computing the log probability contribution for point y from point s.
protected static List<Double>	computeLogKernel(double[] y, List<Event> samples)
protected List<Double>	logPdfRecurse(Event e)	Computing the log probability in a recursive way on the tree.

从类继承的方法 java.lang.Object

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

字段详细资料

numSkipped

public static int numSkipped

构造器详细资料

BallTree

protected BallTree(List<Event> data,
                   int minParent)

Creates a Ball Tree (k-d tree) for the data to reduce the speed of the log probability computation. See: http://www.datalab.uci.edu/papers/kernel_KDD2014.pdf

参数:

data - - The data to create the BallTree for

minParent - - Minimum number of nodes in the area to split to split a region to sub regions.

方法详细资料

logPdfRecurse

protected List<Double> logPdfRecurse(Event e)

Computing the log probability in a recursive way on the tree. We return a list here and not the actual number so we can use the logSumExp trick (Link: ??) to avoid underflow.

参数:

e - - Event to estimate the log probability for

返回:

list of all log probabilities that were computed.

computeLogKernel

protected static List<Double> computeLogKernel(double[] y,
                                               List<Event> samples)

computeLogKernel

protected static double computeLogKernel(double[] y,
                                         double[] s,
                                         double h)

computeLogKernel

protected static double computeLogKernel(double[] y,
                                         double[] s,
                                         double hMax,
                                         double hMin)

The actual math for computing the log probability contribution for point y from point s.

参数:

y - - The point we estimate for.

s - - Observed point in the model

hMax - - The maximum bandwidth in this region

hMin - - The minimum bandwidth in this region

返回:

log probability contribution of 's' to 'y'.