meka.classifiers.multilabel.cc

Class CNode

java.lang.Object

meka.classifiers.multilabel.cc.CNode

All Implemented Interfaces:

Serializable

public class CNode
extends Object
implements Serializable

CNode.java - A Classifier Node class (for CC-like clasifiers).

Version:

June 2013

Author:

Jesse Read

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
CNode(int j, int[] inX, int[] paY) CNode - A Node 'j', taking inputs from all parents inX and paY.

Method Summary

Modifier and Type	Method and Description
void	build(weka.core.Instances D, weka.classifiers.Classifier H) Build - Create transformation for this node, and train classifier of type H upon it.
double	classify(weka.core.Instance x, double[] ypred) Return the argmax on #distribution(Instance, double[]).
double[]	distribution(weka.core.Instance x, double[] ypred) The distribution this this node, given input x.
double[]	distributionT(weka.core.Instance x_) Same as #distribution(Instance, double[]), but the Instance is pre-transformed with ypred inside.
weka.classifiers.Classifier	getClassifier() Returns the underlying classifier.
int[]	getParentsY() getParentsY - get the parents (indices) of this node
static void	main(String[] args) Main - run some tests.
double	sample(weka.core.Instance x, double[] ypred, Random r) Sample the distribution given by #distribution(Instance, double[]).
weka.core.Instance	transform(weka.core.Instance x, double[] ypred) Transform - turn [y1,y2,y3,x1,x2] into [y1,y2,x1,x2].
weka.core.Instances	transform(weka.core.Instances D) Transform - transform dataset D for this node.
static weka.core.Instances	transform(weka.core.Instances D, int c, int[] pa_c) Transform.
void	updateTransform(weka.core.Instance t_, double[] ypred)

Methods inherited from class java.lang.Object

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

Constructor Detail

CNode

public CNode(int j,
             int[] inX,
             int[] paY)

CNode - A Node 'j', taking inputs from all parents inX and paY.

Parameters:

j - the label index of this node

inX - attribute indices going into this node

paY - label indices going into this node

Method Detail

getParentsY

public int[] getParentsY()

getParentsY - get the parents (indices) of this node

transform

public weka.core.Instances transform(weka.core.Instances D)
                              throws Exception

Transform - transform dataset D for this node. this.j defines the current node index, e.g., 3 this.paY[] defines parents, e.g., [1,4] we should remove the rest, e.g., [0,2,5,...,L-1]

Returns:

dataset we should remove all variables from D EXCEPT current node, and parents.

Throws:

Exception

build

public void build(weka.core.Instances D,
                  weka.classifiers.Classifier H)
           throws Exception

Build - Create transformation for this node, and train classifier of type H upon it. The dataset should have class as index 'j', and remove all indices less than L *not* in paY.

Throws:

Exception

distribution

public double[] distribution(weka.core.Instance x,
                             double[] ypred)
                      throws Exception

The distribution this this node, given input x.

Returns:

p( y_j = k | x , y_pred ) for k in {0,1}

Throws:

Exception

distributionT

public double[] distributionT(weka.core.Instance x_)
                       throws Exception

Same as #distribution(Instance, double[]), but the Instance is pre-transformed with ypred inside.

Throws:

Exception

sample

public double sample(weka.core.Instance x,
                     double[] ypred,
                     Random r)
              throws Exception

Sample the distribution given by #distribution(Instance, double[]).

Returns:

y_j ~ p( y_i | x , y_pred )

Throws:

Exception

transform

public weka.core.Instance transform(weka.core.Instance x,
                                    double[] ypred)
                             throws Exception

Transform - turn [y1,y2,y3,x1,x2] into [y1,y2,x1,x2].

Returns:

transformed Instance

Throws:

Exception

updateTransform

public void updateTransform(weka.core.Instance t_,
                            double[] ypred)
                     throws Exception

Throws:

Exception

classify

public double classify(weka.core.Instance x,
                       double[] ypred)
                throws Exception

Return the argmax on #distribution(Instance, double[]).

Returns:

argmax_{k in 0,1,...} p( y_j = k | x , y_pred )

Throws:

Exception

transform

public static weka.core.Instances transform(weka.core.Instances D,
                                            int c,
                                            int[] pa_c)
                                     throws Exception

Transform.

Parameters:

D - original Instances

c - to be the class Attribute

pa_c - the parent indices of c

Returns:

new Instances T

Throws:

Exception

getClassifier

public weka.classifiers.Classifier getClassifier()

Returns the underlying classifier.

Returns:

the classifier

main

public static void main(String[] args)
                 throws Exception

Main - run some tests.

Throws:

Exception