Train a linear model to normalize the components of a dataset to unit variance, and optionally to zero mean. More...

Inheritance diagram for shark::NormalizeComponentsUnitVariance< DataType >:

Public Member Functions
	NormalizeComponentsUnitVariance (bool zeroMean)
	Constructor.

std::string	name () const
	From INameable: return the class name.

void	train (Normalizer< DataType > &model, UnlabeledData< DataType > const &input)

Public Member Functions inherited from shark::AbstractUnsupervisedTrainer< Model >
virtual void	train (ModelType &model, DatasetType const &inputset)=0
	Core of the Trainer interface.

Public Member Functions inherited from shark::INameable
virtual	~INameable ()

Public Member Functions inherited from shark::ISerializable
virtual	~ISerializable ()
	Virtual d'tor.

virtual void	read (InArchive &archive)
	Read the component from the supplied archive.

virtual void	write (OutArchive &archive) const
	Write the component to the supplied archive.

void	load (InArchive &archive, unsigned int version)
	Versioned loading of components, calls read(...).

void	save (OutArchive &archive, unsigned int version) const
	Versioned storing of components, calls write(...).

	BOOST_SERIALIZATION_SPLIT_MEMBER ()

Additional Inherited Members
Public Types inherited from shark::AbstractUnsupervisedTrainer< Model >
typedef Model	ModelType

typedef Model::InputType	InputType

typedef UnlabeledData< InputType >	DatasetType

Detailed Description

template<class DataType = RealVector>
class shark::NormalizeComponentsUnitVariance< DataType >

Train a linear model to normalize the components of a dataset to unit variance, and optionally to zero mean.

: Normalizing the components of a dataset works via training a Normalizer model. This model is then applied to the dataset in order to perform the normalization. The same model can be applied to different datasets.

: The typical use case is that the Normalizer model is trained on the training data. Later, as "test" data comes in, the same model is used, of course without being recalibrated. Thus, the model used for normalization must be independent of the dataset it was trained on.

: Note that subtracting the mean destroys sparsity. Therefore this feature is turned off by default. If you have non-sparse data and you need to move data to zero mean, not only to unit variance, then enable the flag zeroMean in the constructor.

Constructor & Destructor Documentation

template<class DataType = RealVector>

shark::NormalizeComponentsUnitVariance< DataType >::NormalizeComponentsUnitVariance ( bool zeroMean )

inline

Constructor.

: The normalizer scales the data to unit variance. It can also remove the mean of the data. This is usually desired, e.g., for neural network training. Note however that this feature is sometimes undesirable since it can destroy sparsity.

Parameters

zeroMean enable or disable data mean removal