CRFSuite::Trainer Class Reference

The trainer class. More...

List of all members.

Public Member Functions
	Trainer ()
	Construct a trainer.
virtual	~Trainer ()
	Destruct a trainer.
void	clear ()
	Remove all instances in the data set.
void	append (const ItemSequence &xseq, const StringList &yseq, int group)
	Append an instance (item/label sequence) to the data set.
bool	select (const std::string &algorithm, const std::string &type)
	Initialize the training algorithm.
int	train (const std::string &model, int holdout)
	Run the training algorithm.
StringList	params ()
	Obtain the list of parameters.
void	set (const std::string &name, const std::string &value)
	Set a training parameter.
std::string	get (const std::string &name)
	Get the value of a training parameter.
std::string	help (const std::string &name)
	Get the description of a training parameter.
virtual void	message (const std::string &msg)
	Receive messages from the training algorithm.

---

Detailed Description

The trainer class.

This class maintains a data set for training, and provides an interface to various graphical models and training algorithms. The standard procedure for implementing a trainer is:

• create a class by inheriting this class
• overwrite message() function to receive messages of training progress
• call append() to append item/label sequences to the training set
• call select() to specify a graphical model and an algorithm
• call set() to configure parameters specific to the model and algorithm
• call train() to start a training process with the current setting

---

Member Function Documentation

void CRFSuite::Trainer::append	(	const ItemSequence &	xseq,
		const StringList &	yseq,
		int	group
	)

Append an instance (item/label sequence) to the data set.

Parameters:

xseq	The item sequence of the instance.
yseq	The label sequence of the instance. The number of elements in yseq must be identical to that in xseq.
group	The group number of the instance.

Exceptions:

std::invalid_argument	Arguments xseq and yseq are invalid.
std::runtime_error	Out of memory.

std::string CRFSuite::Trainer::get

(

const std::string &

name

)

Get the value of a training parameter.

This function gets a parameter value for the graphical model and training algorithm specified by select() function.

Parameters:

name	The parameter name.

Returns:

std::string The value of the parameter.

Exceptions:

std::invalid_argument

The parameter is not found.

std::string CRFSuite::Trainer::help

(

const std::string &

name

)

Get the description of a training parameter.

This function obtains the help message for the parameter specified by the name. The graphical model and training algorithm must be selected by select() function before calling this function.

Parameters:

name	The parameter name.

Returns:

std::string The description (help message) of the parameter.

virtual void CRFSuite::Trainer::message

(

const std::string &

msg

)

[virtual]

Receive messages from the training algorithm.

Override this member function to receive messages of the training process.

Parameters:

msg	The message

StringList CRFSuite::Trainer::params

(

)

Obtain the list of parameters.

This function returns the list of parameter names available for the graphical model and training algorithm specified by select() function.

Returns:

StringList The list of parameters available for the current graphical model and training algorithm.

bool CRFSuite::Trainer::select	(	const std::string &	algorithm,
		const std::string &	type
	)

Initialize the training algorithm.

Parameters:

algorithm	The name of the training algorithm.
type	The name of the graphical model.

Returns:

bool true if the training algorithm is successfully initialized, false otherwise.

void CRFSuite::Trainer::set	(	const std::string &	name,
		const std::string &	value
	)

Set a training parameter.

This function sets a parameter value for the graphical model and training algorithm specified by select() function.

Parameters:

name	The parameter name.
value	The value of the parameter.

Exceptions:

std::invalid_argument

The parameter is not found.

int CRFSuite::Trainer::train	(	const std::string &	model,
		int	holdout
	)

Run the training algorithm.

This function starts the training algorithm with the data set given by append() function. After starting the training process, the training algorithm invokes the virtual function message() to report the progress of the training process.

Parameters:

model	The filename to which the trained model is stored. If this value is empty, this function does not write out a model file.
holdout	The group number of holdout evaluation. The instances with this group number will not be used for training, but for holdout evaluation. Specify -1 to use all instances for training.

Returns:

int The status code.