Class DisjunctMatrixValidationOptions

java.lang.Object

com.knezevic.edaf.v3.problems.discrete.disjunct.DisjunctMatrixValidationOptions

public final class DisjunctMatrixValidationOptions
extends Object

Controls when validation uses exhaustive enumeration versus statistically justified sampling, and configures the associated confidence bound.

Constructor Summary

Constructors

Constructor	Description
DisjunctMatrixValidationOptions(long maxExactSubsets, long sampleSize, double confidenceLevel, double absoluteError, long randomSeed)	Creates custom options.

Method Summary

Modifier and Type	Method	Description
double	absoluteError()
double	confidenceLevel()
static DisjunctMatrixValidationOptions	defaults()	Reasonable defaults for practical validation: exact for up to 200k subsets, otherwise sample via Hoeffding target.
long	maxExactSubsets()
long	randomSeed()
long	resolvedSampleSize()	Resolves effective sample size.
long	sampleSize()

Methods inherited from class java.lang.Object

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

Constructor Details

DisjunctMatrixValidationOptions

public DisjunctMatrixValidationOptions(long maxExactSubsets,
 long sampleSize,
 double confidenceLevel,
 double absoluteError,
 long randomSeed)

Creates custom options.

Parameters:

maxExactSubsets - max number of t-subsets for exact mode.

sampleSize - explicit sample size; use 0 to derive from confidence/error.

confidenceLevel - confidence in (0,1), e.g. 0.95.

absoluteError - target Hoeffding half-width, e.g. 0.02.

randomSeed - sampling seed for deterministic reproducibility.

Method Details

defaults

public static DisjunctMatrixValidationOptions defaults()

Reasonable defaults for practical validation: exact for up to 200k subsets, otherwise sample via Hoeffding target.

maxExactSubsets

public long maxExactSubsets()

sampleSize

public long sampleSize()

confidenceLevel

public double confidenceLevel()

absoluteError

public double absoluteError()

randomSeed

public long randomSeed()

resolvedSampleSize

public long resolvedSampleSize()

Resolves effective sample size. If explicit sample size is not supplied, it is derived from Hoeffding inequality: n >= ln(2/alpha)/(2*eps^2) where alpha = 1 - confidence.