Package | Description |
---|---|
de.jstacs.classifiers.differentiableSequenceScoreBased.sampling |
Provides the classes for
AbstractScoreBasedClassifier s that are based on
SamplingDifferentiableStatisticalModel s
and that sample parameters using the Metropolis-Hastings algorithm. |
Modifier and Type | Method and Description |
---|---|
protected SamplingScoreBasedClassifier.DiffSMSamplingComponent |
SamplingScoreBasedClassifier.getSamplingComponent()
Returns a sampling component suited for this
SamplingScoreBasedClassifier |
Modifier and Type | Method and Description |
---|---|
protected void |
SamplingScoreBasedClassifier.precomputeBurnInLength(SamplingScoreBasedClassifier.DiffSMSamplingComponent sfsc)
Precomputes the length of the burn-in phase, e.g.
|
protected void |
SamplingScoreBasedClassifier.sample(SamplingScoreBasedClassifier.DiffSMSamplingComponent sfsc,
Function function)
Samples as many steps as needed to get into the stationary phase according to
SamplingScoreBasedClassifier.burnInTest and then samples the number of
stationary steps as set in SamplingScoreBasedClassifier.params . |
protected double |
SamplingScoreBasedClassifier.sampleNSteps(Function function,
SamplingScoreBasedClassifier.DiffSMSamplingComponent component,
BurnInTest test,
int numSteps,
SamplingScoreBasedClassifier.SamplingScheme scheme)
Samples a predefined number of steps appended to the current sampling
|