| Package | Description |
|---|---|
| de.jstacs.classifiers.differentiableSequenceScoreBased |
Provides the classes for
Classifiers that are based on SequenceScores.It includes a sub-package for discriminative objective functions, namely conditional likelihood and supervised posterior, and a separate sub-package for the parameter priors, that can be used for the supervised posterior. |
| de.jstacs.classifiers.differentiableSequenceScoreBased.gendismix |
Provides an implementation of a classifier that allows to train the parameters of a set of
DifferentiableStatisticalModels by
a unified generative-discriminative learning principle. |
| de.jstacs.classifiers.differentiableSequenceScoreBased.logPrior |
Provides a general definition of a parameter log-prior and a number of implementations of Laplace and Gaussian priors.
|
| de.jstacs.classifiers.differentiableSequenceScoreBased.msp |
Provides an implementation of a classifier that allows to train the parameters of a set of
DifferentiableStatisticalModels either
by maximum supervised posterior (MSP) or by maximum conditional likelihood (MCL). |
| de.jstacs.classifiers.differentiableSequenceScoreBased.sampling |
Provides the classes for
AbstractScoreBasedClassifiers that are based on
SamplingDifferentiableStatisticalModels
and that sample parameters using the Metropolis-Hastings algorithm. |
| de.jstacs.sequenceScores.statisticalModels.trainable |
Provides all
TrainableStatisticalModels, which can
be learned from a single DataSet. |
| Class and Description |
|---|
| LogPrior
The abstract class for any log-prior used e.g.
|
| Class and Description |
|---|
| LogPrior
The abstract class for any log-prior used e.g.
|
| Class and Description |
|---|
| CompositeLogPrior
This class implements a composite prior that can be used for DifferentiableStatisticalModel.
|
| DoesNothingLogPrior
This class defines a
LogPrior that does not penalize any parameter. |
| LogPrior
The abstract class for any log-prior used e.g.
|
| SeparateLogPrior
Abstract class for priors that penalize each parameter value independently
and have some variances (and possible means) as hyperparameters.
|
| SimpleGaussianSumLogPrior
This class implements a prior that is a product of Gaussian distributions
with mean 0 and equal variance for each parameter.
|
| Class and Description |
|---|
| LogPrior
The abstract class for any log-prior used e.g.
|
| Class and Description |
|---|
| LogPrior
The abstract class for any log-prior used e.g.
|
| Class and Description |
|---|
| LogPrior
The abstract class for any log-prior used e.g.
|