PhasedPMarkov Class Reference

Phased Parcimonious Markov modelling, estimation and simulation. More...

#include <seqpp/PhasedPMarkov.h>

Inheritance diagram for PhasedPMarkov:

List of all members.

Public Member Functions
	PhasedPMarkov (Partition &part, const SequenceSet &seqset, short phase, short initial_phase=0, const string &prior_alpha_file=string(), bool motif_prior=true, double penalty=0., const string &xmlfile=string())
	Constructor 1 from a SequenceSet.
	PhasedPMarkov (Partition &part, const Sequence &seq, short phase, short initial_phase=0, const string &prior_alpha_file=string(), bool motif_prior=true, double penalty=0., const Translator &trans=Translator(), const string &xmlfile=string())
	Constructor 2 from a Sequence.
	PhasedPMarkov (Partition &part, unsigned long **count, short size, short order, short phase, const string &prior_alpha_file=string(), bool motif_prior=true, double penalty=0., const Translator &trans=Translator(), const string &xmlfile=string())
	Constructor 3 from a coded-word count.
	PhasedPMarkov (Partition &part, const string &count_file, short size, short order, short phase, const string &prior_alpha_file=string(), bool motif_prior=true, double penalty=0., const Translator &trans=Translator(), const string &xmlfile=string())
	Constructor 4 from a file containing coded-word count.
	PhasedPMarkov (Partition &part, short size, short order, short phase, const string &prior_alpha_file=string(), bool motif_prior=true, double penalty=0., bool alloc=true)
	Basic Constructor 5. No estimation.
void	select (unsigned long **count, bool decal_required, const Translator &trans=Translator(), const string &xmlfile=string())
	performs the A Posteriori Maximisation
template<class TSeq1, class TSeq2>
double	mean_post_log_likelihood (const TSeq1 &tseq_train, const TSeq2 &tseq_eval, short initial_phase_train=0, short initial_phase_eval=0)
double	mean_post_log_likelihood (unsigned long **count_train, bool decal_required_t, unsigned long **count_eval, bool decal_required_e)
	return the mean post likelihood over the parameters and the trees
template<class TSeq>
double	mean_post_log_likelihood (const TSeq &tseq_eval, short initial_phase_eval=0)
double	mean_post_log_likelihood (unsigned long **count_eval, bool decal_required_e)
	return the mean post likelihood over the parameters and the trees
double	mean_post_log_likelihood ()
	return the mean post likelihood over the parameters and the trees
template<class TSeq>
double	post_log_likelihood (const TSeq &tseq_eval, short initial_phase_eval=0)
	compute the mean posterior likelihood over the parameters
double	post_log_likelihood (unsigned long **count_eval, bool decal_required_e)
	compute the mean posterior likelihood over the parameters
void	draw (unsigned long **count, bool decal_required, gsl_rng *r, const Translator &trans=Translator(), const string &xmlfile=string())
	draws a model
void	draw (gsl_rng *r, const Translator &trans=Translator(), const string &xmlfile=string())
	draws a model
void	info_nb_leaves () const
	returns info on the number of leaves for each phase
	~PhasedPMarkov ()
	Destructor.
Protected Attributes
vector< pmm_forest * >	_p_f
	Parcimonious Context Trees.

---

Detailed Description

Phased Parcimonious Markov modelling, estimation and simulation.

PhasedPMarkov is a PhasedMarkov object with a different estimation step. This object performs the estimation with the Parcimonious Markov algorithm and then transfoms, once per phase, the parcimonious context tree in a markovian matrix. xml outputs can be activated to save the associated trees.

---

Constructor & Destructor Documentation

PhasedPMarkov::PhasedPMarkov	(	Partition &	part,
		const SequenceSet &	seqset,
		short	phase,
		short	initial_phase = 0,
		const string &	prior_alpha_file = string(),
		bool	motif_prior = true,
		double	penalty = 0.,
		const string &	xmlfile = string()
	)			[inline]

Constructor 1 from a SequenceSet.

Parameters:

	part	associated partition
	seqset	a set of sequences for estimation
	phase	phase
	initial_phase	phase of the first element of each sequence
	prior_alpha_file	file containing the alpha for the a priori law, one value per alphabet element, and for each phase (separated by a "#Phase i")
	motif_prior	activates a weight-function on the priors, weight proportionnal on each motif in the tree
	penalty	penalty on the leaves number, by default 0
	xmlfile	xmlfile for tree saving (if xml2 activated)

References PhasedMarkov::_order, _p_f, PhasedMarkov::_phase, PhasedMarkov::_prior_alpha, PhasedMarkov::_size, PrimaryCount::count_p_occurencies(), PrimaryCount::get_p_count(), PrimarySequenceSet< TSequence >::get_translator(), and select().

PhasedPMarkov::PhasedPMarkov	(	Partition &	part,
		const Sequence &	seq,
		short	phase,
		short	initial_phase = 0,
		const string &	prior_alpha_file = string(),
		bool	motif_prior = true,
		double	penalty = 0.,
		const Translator &	trans = Translator(),
		const string &	xmlfile = string()
	)			[inline]

Constructor 2 from a Sequence.

Parameters:

	part	associated partition
	seq	sequence for estimation
	phase	phase
	initial_phase	phase of the first element of each sequence
	prior_alpha_file	file containing the alpha for the a priori law, one value per alphabet element, and for each phase (separated by a "#Phase i")
	motif_prior	activates a weight-function on the priors, weight proportionnal on each motif in the tree
	penalty	penalty on the leaves number, by default 0
	trans	a Translator is required only for the xml saving
	xmlfile	xmlfile for tree saving (if xml2 activated)

References PhasedMarkov::_order, _p_f, PhasedMarkov::_phase, PhasedMarkov::_prior_alpha, PhasedMarkov::_size, PrimaryCount::count_p_occurencies(), PrimaryCount::get_p_count(), and select().

PhasedPMarkov::PhasedPMarkov	(	Partition &	part,
		unsigned long **	count,
		short	size,
		short	order,
		short	phase,
		const string &	prior_alpha_file = string(),
		bool	motif_prior = true,
		double	penalty = 0.,
		const Translator &	trans = Translator(),
		const string &	xmlfile = string()
	)			[inline]

Constructor 3 from a coded-word count.

Parameters:

	part	associated partition
	count	count of all the coded word(base size) of size order+1 for each phase, for estimation
	size	alphabet size
	order	markovian order associated to the word count
	phase	phase
	prior_alpha_file	file containing the alpha for the a priori law, one value per alphabet element, and for each phase (separated by a "#Phase i")
	motif_prior	activates a weight-function on the priors, weight proportionnal on each motif in the tree
	penalty	penalty on the leaves number, by default 0
	trans	a Translator is required only for the xml saving
	xmlfile	xmlfile for tree saving (if xml2 activated)

References PhasedMarkov::_order, _p_f, PhasedMarkov::_phase, PhasedMarkov::_prior_alpha, PhasedMarkov::_size, and select().

PhasedPMarkov::PhasedPMarkov	(	Partition &	part,
		const string &	count_file,
		short	size,
		short	order,
		short	phase,
		const string &	prior_alpha_file = string(),
		bool	motif_prior = true,
		double	penalty = 0.,
		const Translator &	trans = Translator(),
		const string &	xmlfile = string()
	)			[inline]

Constructor 4 from a file containing coded-word count.

Parameters:

	part	associated partition
	count_file	file with count
	size	alphabet size
	order	markovian order associated to the word count
	phase	phase
	prior_alpha_file	file containing the alpha for the a priori law, one value per alphabet element, and for each phase (separated by a "#Phase i")
	motif_prior	activates a weight-function on the priors, weight proportionnal on each motif in the tree
	penalty	penalty on the leaves number, by default 0
	trans	a Translator is required only for the xml saving
	xmlfile	xmlfile for tree saving (if xml2 activated)

References PhasedMarkov::_nPi, PhasedMarkov::_order, _p_f, PhasedMarkov::_phase, PhasedMarkov::_prior_alpha, PhasedMarkov::_size, and select().

PhasedPMarkov::PhasedPMarkov	(	Partition &	part,
		short	size,
		short	order,
		short	phase,
		const string &	prior_alpha_file = string(),
		bool	motif_prior = true,
		double	penalty = 0.,
		bool	alloc = true
	)			[inline]

Basic Constructor 5. No estimation.

Parameters:

	part	associated partition
	size	alphabet size
	order	markovian order associated to the word count
	phase	phase
	prior_alpha_file	file containing the alpha for the a priori law, one value per alphabet element, and for each phase (separated by a "#Phase i")
	motif_prior	activates a weight-function on the priors, weight proportionnal on each motif in the tree
	penalty	penalty on the leaves number, by default 0
	alloc	true for matrices memory allocation

References PhasedMarkov::_order, _p_f, PhasedMarkov::_phase, PhasedMarkov::_prior_alpha, and PhasedMarkov::_size.

---

Member Function Documentation

void PhasedPMarkov::select	(	unsigned long **	count,
		bool	decal_required,
		const Translator &	trans = Translator(),
		const string &	xmlfile = string()
	)			[inline]

performs the A Posteriori Maximisation

Parameters:

	count	count of all the coded word(base size) of size order+1 for each phase, for estimation
	decal_required	necessary when using a count of word from 1-word to (_max_depth+1)-word
	trans	a Translator is required for xml tree saving (if xml2 activated)
	xmlfile	xmlfile for tree saving (if xml2 activated)

References PhasedMarkov::_jump, PhasedMarkov::_nb_param, _p_f, PhasedMarkov::_phase, PhasedMarkov::_Pis, and PhasedMarkov::_size.

Referenced by PhasedPMarkov().

template<class TSeq1, class TSeq2>

double PhasedPMarkov::mean_post_log_likelihood	(	const TSeq1 &	tseq_train,
		const TSeq2 &	tseq_eval,
		short	initial_phase_train = 0,
		short	initial_phase_eval = 0
	)			[inline]

Parameters:

	tseq_train	sequence(s) (set) for the training step
	initial_phase_train	phase of the first element of each sequence
	tseq_eval	sequence(s) (set) for the evaluation step
	initial_phase_eval	phase of the first element of each sequence

References PhasedMarkov::_phase, and mean_post_log_likelihood().

double PhasedPMarkov::mean_post_log_likelihood	(	unsigned long **	count_train,
		bool	decal_required_t,
		unsigned long **	count_eval,
		bool	decal_required_e
	)			[inline]

return the mean post likelihood over the parameters and the trees

Parameters:

	count_train	count of all the coded word(base size) of size order+1 for each phase, for training step
	decal_required_t	necessary when using a count of word from 1-word to (_max_depth+1)-word
	count_eval	count of all the coded word(base size) of size order+1 for each phase, for evaluation step
	decal_required_e	necessary when using a count of word from 1-word to (_max_depth+1)-word

References PhasedMarkov::_jump, _p_f, and PhasedMarkov::_phase.

template<class TSeq>

double PhasedPMarkov::mean_post_log_likelihood	(	const TSeq &	tseq_eval,
		short	initial_phase_eval = 0
	)			[inline]

Parameters:

	tseq_eval	sequence(s) (set) for the evaluation step
	initial_phase_eval	phase of the first element of each sequence

References PhasedMarkov::_phase, and mean_post_log_likelihood().

double PhasedPMarkov::mean_post_log_likelihood	(	unsigned long **	count_eval,
		bool	decal_required_e
	)			[inline]

return the mean post likelihood over the parameters and the trees

Parameters:

	count_eval	count of all the coded word(base size) of size order+1 for each phase, for evaluation step
	decal_required_e	necessary when using a count of word from 1-word to (_max_depth+1)-word

References PhasedMarkov::_jump, _p_f, and PhasedMarkov::_phase.

template<class TSeq>

double PhasedPMarkov::post_log_likelihood	(	const TSeq &	tseq_eval,
		short	initial_phase_eval = 0
	)			[inline]

compute the mean posterior likelihood over the parameters

Parameters:

	tseq_eval	sequence(s) (set) for the evaluation step
	initial_phase_eval	phase of the first element of each sequence

References PhasedMarkov::_order, and PhasedMarkov::_phase.

double PhasedPMarkov::post_log_likelihood	(	unsigned long **	count_eval,
		bool	decal_required_e
	)			[inline]

compute the mean posterior likelihood over the parameters

Parameters:

	count_eval	count of all the coded word(base size) of size order+1 for each phase, for evaluation step
	decal_required_e	necessary when using a count of word from 1-word to (_max_depth+1)-word

References PhasedMarkov::_jump, _p_f, and PhasedMarkov::_phase.

void PhasedPMarkov::draw	(	unsigned long **	count,
		bool	decal_required,
		gsl_rng *	r,
		const Translator &	trans = Translator(),
		const string &	xmlfile = string()
	)			[inline]

draws a model

Parameters:

	count	count of all the coded word(base size) of size order+1 for each phase, for estimation
	decal_required	necessary when using a count of word from 1-word to (_max_depth+1)-word
	r	gsl random generator
	trans	a Translator is required for xml tree saving (if xml2 activated)
	xmlfile	xmlfile for tree saving (if xml2 activated)

References PhasedMarkov::_jump, PhasedMarkov::_Mus, PhasedMarkov::_nb_param, _p_f, PhasedMarkov::_phase, PhasedMarkov::_Pis, PhasedMarkov::_size, and PhasedMarkov::compute_stat_laws().

void PhasedPMarkov::draw	(	gsl_rng *	r,
		const Translator &	trans = Translator(),
		const string &	xmlfile = string()
	)			[inline]

draws a model

Parameters:

	r	random generator
	trans	a Translator is required for xml tree saving (if xml2 activated)
	xmlfile	xmlfile for tree saving (if xml2 activated)

References PhasedMarkov::_Mus, PhasedMarkov::_nb_param, _p_f, PhasedMarkov::_phase, PhasedMarkov::_Pis, PhasedMarkov::_size, and PhasedMarkov::compute_stat_laws().

---

The documentation for this class was generated from the following file:

• PhasedPMarkov.h