Package edu.wisc.game.math

Class MannWhitneyComparison

java.lang.Object

edu.wisc.game.math.MannWhitneyComparison

public class MannWhitneyComparison
extends Object

Comparing players or rules based on the Mann-Whitney test

Nested Class Summary

Nested Classes

Modifier and Type	Class	Description
static enum	MannWhitneyComparison.Mode

Constructor Summary

Constructors

Constructor	Description
MannWhitneyComparison(MannWhitneyComparison.Mode _mode)

Method Summary

Modifier and Type	Method	Description
String	doCompare(String nickname, String rule, Comparandum[][] allComp, Fmter fm, File[] csvOut)	Carries out the comparison of the performance for different "keys" (algo nicknames or rule sets).
static File[]	expandCsvOutDir(File csvOutDir)	Suggests names for CSV output files
static void	main(String[] argv)
Comparandum[][]	mkMlcComparanda(String nickname, String rule)	Creates a list of comparanda based on MLC data, either to compare ML algos or to compare rule sets.

Methods inherited from class java.lang.Object

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

Constructor Details

MannWhitneyComparison

public MannWhitneyComparison(MannWhitneyComparison.Mode _mode)

Parameters:

_mode - What are we going to compare? Rules (by the algos' performance on them), rules (by the humans' performance on them) or ML algos (by their performance on the

Method Details

mkMlcComparanda

public Comparandum[][] mkMlcComparanda(String nickname,
 String rule)

Creates a list of comparanda based on MLC data, either to compare ML algos or to compare rule sets.

Returns:

{learnedOnes[], nonLearnedOnes[]}

doCompare

public String doCompare(String nickname,
 String rule,
 Comparandum[][] allComp,
 Fmter fm,
 File[] csvOut)

Carries out the comparison of the performance for different "keys" (algo nicknames or rule sets). In the CMP_ALGOS mode, a particular rule set is chosen, and ML algorithms are ranked by their performance on that rule set; thus the "key" is the algo nickname. In the CMP_RULES mode, a particular algorithm is chosen (the nickname) is specified, and the rule sets are ranked by their ease for this algorithm; thus the rule sets names are keys.

Parameters:

allComp - The things to compare. allComp[0] is the list of "learned" comparanda, and allComp[1] is the list of unlearned ones. The comparison is done under different criteria in each group.

csvOut - If non-null, designates 3 files into which the raw M-W matrix, the M-W ratio matrix, and the final table will be written into

main

public static void main(String[] argv)

expandCsvOutDir

public static File[] expandCsvOutDir(File csvOutDir)

Suggests names for CSV output files