meka.core

Class PSUtils

java.lang.Object

meka.core.PSUtils

public abstract class PSUtils
extends Object

PSUtils.java - Handy Utils for working with Pruned Sets. Essentially, we have a P parameter for pruning and an N parameter for reintroduction.

Version:

June 2014

Author:

Jesse Read

Constructor Summary

Constructors

Constructor and Description
PSUtils()

Method Summary

Modifier and Type	Method and Description
static LabelSet[]	convert(LabelSet[] sparseY, HashMap<LabelSet,Integer> map) Given N labelsets 'sparseY', use a count 'map' to
static double[]	convertDistribution(double[] p, int L, weka.core.Instances iTemplate) Deprecated.
static double[]	convertDistribution(double[] p, int L, LabelSet[] meta_labels) Convert Distribution - Given the posterior across combinations, return the distribution across labels.
static weka.core.Instance	convertInstance(weka.core.Instance x, int L, weka.core.Instances template) Convert a multi-label instance into a multi-class instance, according to a template.
static HashMap<LabelSet,Integer>	countCombinationsSparse(weka.core.Instances D, int L) CountCombinationsSparse - return a mapping of each distinct label combination and its count.
static HashMap<LabelSet,Integer>	countCombinationsSparseSubset(weka.core.Instances D, int[] indices) CountCombinationsSparseSubset - like CountCombinationsSparse, but only interested in 'indices[]' wrt 'D'.
static int	countSubsets(LabelSet ysub, Set<LabelSet> Y) Count Subsets - returns the number of times labelset 'ysub' exists as a subset in 'Y'.
static LabelSet[]	cover(LabelSet y, HashMap<LabelSet,Integer> map) Cover - cover 'y' completely (or as best as possible) with sets from 'map'.
static Set<LabelSet>	cover(LabelSet y, SortedSet<LabelSet> S, Comparator cmp)
static LabelSet[]	getAllSubsets(LabelSet y, HashMap<LabelSet,Integer> map) GetAllSubsets - Get all frequent subsets of 'y' according to 'map'.
static SortedSet<LabelSet>	getSortedSubsets(LabelSet y, HashMap<LabelSet,Integer> map) Get Sorted Subsets - get all subsets of 'y' in the set 'set'; sorted according to length, and counts in 'map'.
static SortedSet<LabelSet>	getSortedSubsets(LabelSet y, Set<LabelSet> set, Comparator cmp) Get Sorted Subsets - get all subsets of 'y' in the set 'set'; sorted according to 'cmp'.
static Set<LabelSet>	getSubsets(LabelSet y, Set<LabelSet> set) Get Subsets - get all subsets of 'y' in the set 'set'.
static LabelSet[]	getTopNSubsets(LabelSet y, HashMap<LabelSet,Integer> map, int n) GetTopNSubsets - Don't cover all (like cover(y,map), rather only the top 'n')
static SortedSet<LabelSet>	getTopNSubsetsAsSet(LabelSet y, HashMap<LabelSet,Integer> map, int n)
static LabelSet	getTopSubset(LabelSet y, HashMap<LabelSet,Integer> map)
static weka.core.Instances	LCTransformation(weka.core.Instances D)
static weka.core.Instances	LCTransformation(weka.core.Instances D, int L)
static HashMap<LabelSet,Integer>	loadMap(String filename) LoadMap - Load the HashMap stored in 'filename'.
static LabelSet[]	makeLabelSetMap(weka.core.Instances T)
static weka.core.Instance[]	PSTransformation(weka.core.Instance x, int L, HashMap<LabelSet,Integer> map, int n, weka.core.Instances iTemplate) Transform one instance into multi-class representations (an array of possibly multiple single-label instances).
static weka.core.Instances	PSTransformation(weka.core.Instances D, int P, int N)
static weka.core.Instances	PSTransformation(weka.core.Instances D, int L, int P, int N)
static weka.core.Instances	PSTransformation(weka.core.Instances D, int L, String cname, int p, int n) Transform instances into a multi-class representation.
static double[]	recombination_t(double[] p, int L, weka.core.Instances iTemplate) Convert Distribution - Given the posterior across combinations, return the distribution across labels.
static double[]	recombination_t(double[] p, int L, LabelSet[] map)
static double[]	recombination(double[] p, int L, LabelSet[] map) Convert Distribution - Given the posterior across combinations, return the distribution across labels.
static void	saveMap(String filename, HashMap<LabelSet,Integer> map) SaveMap - Save the HashMap 'map' to the file 'filename'.
static weka.core.Instances	SLTransformation(weka.core.Instances D, int L, String cname, int p, int n) Transform instances into a multi-class representation.
static int	sumCounts(HashMap<LabelSet,Integer> map) Sum Counts - sum all the values in 'map'.

Methods inherited from class java.lang.Object

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

Constructor Detail

PSUtils

public PSUtils()

Method Detail

sumCounts

public static int sumCounts(HashMap<LabelSet,Integer> map)

Sum Counts - sum all the values in 'map'.

countSubsets

public static int countSubsets(LabelSet ysub,
                               Set<LabelSet> Y)

Count Subsets - returns the number of times labelset 'ysub' exists as a subset in 'Y'.

getSubsets

public static Set<LabelSet> getSubsets(LabelSet y,
                                       Set<LabelSet> set)

Get Subsets - get all subsets of 'y' in the set 'set'.

getSortedSubsets

public static SortedSet<LabelSet> getSortedSubsets(LabelSet y,
                                                   Set<LabelSet> set,
                                                   Comparator cmp)

Get Sorted Subsets - get all subsets of 'y' in the set 'set'; sorted according to 'cmp'.

getSortedSubsets

public static SortedSet<LabelSet> getSortedSubsets(LabelSet y,
                                                   HashMap<LabelSet,Integer> map)

Get Sorted Subsets - get all subsets of 'y' in the set 'set'; sorted according to length, and counts in 'map'.

cover

public static LabelSet[] cover(LabelSet y,
                               HashMap<LabelSet,Integer> map)

Cover - cover 'y' completely (or as best as possible) with sets from 'map'.

Parameters:

y - a LabelSet, e.g., [0,2,7]

map - a map of LabelSets to counts e.g., {[0,2,7]:39,...}

Returns:

the sets to cover y (or just y, if it already covers itself).

cover

public static Set<LabelSet> cover(LabelSet y,
                                  SortedSet<LabelSet> S,
                                  Comparator cmp)

getAllSubsets

public static LabelSet[] getAllSubsets(LabelSet y,
                                       HashMap<LabelSet,Integer> map)

GetAllSubsets - Get all frequent subsets of 'y' according to 'map'.

Parameters:

y - a labelset, e.g., [0,2,7]

map - a map of labelsets to counts e.g., {[0,2]:39, [2,7]:5, [2,9]:24...}

Returns:

the LabelSets to use to decompose y into, e.g., [[0,2],[2,7]]

getTopNSubsets

public static LabelSet[] getTopNSubsets(LabelSet y,
                                        HashMap<LabelSet,Integer> map,
                                        int n)

GetTopNSubsets - Don't cover all (like cover(y,map), rather only the top 'n')

Parameters:

y - a labelset, e.g., [0,2,7]

map - a map of labelsets to counts e.g., {[0,2]:39, [2,7]:5, [2,9]:24...}

n - the number of sets to take

Returns:

the LabelSets to use to decompose y into, e.g., [[0,2],[2,7]]

getTopNSubsetsAsSet

public static SortedSet<LabelSet> getTopNSubsetsAsSet(LabelSet y,
                                                      HashMap<LabelSet,Integer> map,
                                                      int n)

getTopSubset

public static LabelSet getTopSubset(LabelSet y,
                                    HashMap<LabelSet,Integer> map)

countCombinationsSparseSubset

public static HashMap<LabelSet,Integer> countCombinationsSparseSubset(weka.core.Instances D,
                                                                      int[] indices)

CountCombinationsSparseSubset - like CountCombinationsSparse, but only interested in 'indices[]' wrt 'D'.

Parameters:

D - dataset

indices - indices we are interested in

Returns:

a HashMap where a LabelSet representation of each label combination is associated with an Integer count, e.g., [3,7,14],3

countCombinationsSparse

public static final HashMap<LabelSet,Integer> countCombinationsSparse(weka.core.Instances D,
                                                                      int L)

CountCombinationsSparse - return a mapping of each distinct label combination and its count.

Parameters:

D - dataset

L - number of labels

Returns:

a HashMap where a LabelSet representation of each label combination is associated with an Integer count, e.g., [3,7,14],3

convertDistribution

@Deprecated
public static double[] convertDistribution(double[] p,
                                                       int L,
                                                       weka.core.Instances iTemplate)

Deprecated.

Convert Distribution - Given the posterior across combinations, return the distribution across labels.
TODO Use recombination!!!

Parameters:

p - the posterior of the super classes (combinations), e.g., P([1,3],[2]) = [1,0]

L - the number of labels

Returns:

the distribution across labels, e.g., P(1,2,3) = [1,0,1]

See Also:

recombination(double[],int,LabelSet[])

convertDistribution

public static double[] convertDistribution(double[] p,
                                           int L,
                                           LabelSet[] meta_labels)

Convert Distribution - Given the posterior across combinations, return the distribution across labels.

Parameters:

p - the posterior of the super classes (combinations), e.g., P([1,3],[2]) = [0.3,0.7]

L - the number of labels, e.g., L = 3

meta_labels - typical mapping, e.g., [13] to [1,3]

Returns:

the distribution across labels, e.g., P(1,2,3) = [0.3,0.7,0.3]

makeLabelSetMap

public static final LabelSet[] makeLabelSetMap(weka.core.Instances T)

recombination

public static final double[] recombination(double[] p,
                                           int L,
                                           LabelSet[] map)

Convert Distribution - Given the posterior across combinations, return the distribution across labels.

Parameters:

p - the posterior of the super classes (combinations), e.g., P([1,3],[2]) = [1,0]

L - the number of labels

Returns:

the distribution across labels, e.g., P(1,2,3) = [1,0,1]

recombination_t

public static final double[] recombination_t(double[] p,
                                             int L,
                                             weka.core.Instances iTemplate)

Convert Distribution - Given the posterior across combinations, return the distribution across labels.

Parameters:

p - the posterior of the super classes (combinations), e.g., P([1,3],[2]) = [0.3,0.7]

L - the number of labels

Returns:

the distribution across labels, e.g., P(1,2,3) = [0.3,0.7,0.3]

recombination_t

public static final double[] recombination_t(double[] p,
                                             int L,
                                             LabelSet[] map)

convertInstance

public static weka.core.Instance convertInstance(weka.core.Instance x,
                                                 int L,
                                                 weka.core.Instances template)

Convert a multi-label instance into a multi-class instance, according to a template.

LCTransformation

public static weka.core.Instances LCTransformation(weka.core.Instances D)

LCTransformation

public static weka.core.Instances LCTransformation(weka.core.Instances D,
                                                   int L)

PSTransformation

public static weka.core.Instances PSTransformation(weka.core.Instances D,
                                                   int P,
                                                   int N)

PSTransformation

public static weka.core.Instances PSTransformation(weka.core.Instances D,
                                                   int L,
                                                   int P,
                                                   int N)

PSTransformation

public static weka.core.Instances PSTransformation(weka.core.Instances D,
                                                   int L,
                                                   String cname,
                                                   int p,
                                                   int n)

Transform instances into a multi-class representation.

Parameters:

D - original dataset

L - number of labels in the original dataset

cname - class name for the new dataset (may want to encode the list of indices here for RAkEL-like methods)

p - pruning value

n - restoration value

Returns:

transformed dataset

PSTransformation

public static weka.core.Instance[] PSTransformation(weka.core.Instance x,
                                                    int L,
                                                    HashMap<LabelSet,Integer> map,
                                                    int n,
                                                    weka.core.Instances iTemplate)

Transform one instance into multi-class representations (an array of possibly multiple single-label instances).

Parameters:

x - instance

L - number of labels in the instance

map - a map of labelsets to their frequencies

n - restoration value

Returns:

transformed instances

SLTransformation

public static weka.core.Instances SLTransformation(weka.core.Instances D,
                                                   int L,
                                                   String cname,
                                                   int p,
                                                   int n)

Transform instances into a multi-class representation.

Parameters:

D - original dataset

L - number of labels in that dataset

cname - class name for the new dataset (may want to encode the list of indices here for RAkEL-like methods)

p - pruning value

n - restoration value

Returns:

transformed dataset

convert

public static final LabelSet[] convert(LabelSet[] sparseY,
                                       HashMap<LabelSet,Integer> map)

Given N labelsets 'sparseY', use a count 'map' to

saveMap

public static final void saveMap(String filename,
                                 HashMap<LabelSet,Integer> map)
                          throws Exception

SaveMap - Save the HashMap 'map' to the file 'filename'.

Throws:

Exception

loadMap

public static HashMap<LabelSet,Integer> loadMap(String filename)
                                         throws Exception

LoadMap - Load the HashMap stored in 'filename'.

Throws:

Exception