Detailed Description

The class LinearHMM is for learning Higher Order Markov chains.

While learning the parameters ${\bf \theta}$ in

$\begin{eqnarray*} P({\bf x}|{\bf \theta}^\pm)&=&P(x_1, \ldots, x_N|{\bf \theta}^\pm)\\ &=&P(x_1,\ldots,x_{d}|{\bf \theta}^\pm)\prod_{i=d+1}^N P(x_i|x_{i-1},\ldots,x_{i-d},{\bf \theta}^\pm) \end{eqnarray*}$

are determined.

A more detailed description can be found in

Durbin et.al, Biological Sequence Analysis -Probabilistic Models of Proteins and Nucleic Acids, 1998

Definition at line 39 of file LinearHMM.h.

Inheritance diagram for CLinearHMM:

[legend]

Public Member Functions
	CLinearHMM ()
	CLinearHMM (CStringFeatures< uint16_t > *f)
	CLinearHMM (int32_t p_num_features, int32_t p_num_symbols)
virtual	~CLinearHMM ()
virtual bool	train (CFeatures *data=NULL)
bool	train (const int32_t *indizes, int32_t num_indizes, float64_t pseudo_count)
float64_t	get_log_likelihood_example (uint16_t *vector, int32_t len)
float64_t	get_likelihood_example (uint16_t *vector, int32_t len)
virtual float64_t	get_log_likelihood_example (int32_t num_example)
virtual float64_t	get_log_derivative (int32_t num_param, int32_t num_example)
virtual float64_t	get_log_derivative_obsolete (uint16_t obs, int32_t pos)
virtual float64_t	get_derivative_obsolete (uint16_t *vector, int32_t len, int32_t pos)
virtual int32_t	get_sequence_length ()
virtual int32_t	get_num_symbols ()
virtual int32_t	get_num_model_parameters ()
virtual float64_t	get_positional_log_parameter (uint16_t obs, int32_t position)
virtual float64_t	get_log_model_parameter (int32_t num_param)
virtual SGVector< float64_t >	get_log_transition_probs ()
virtual bool	set_log_transition_probs (SGVector< float64_t > probs)
virtual SGVector< float64_t >	get_transition_probs ()
virtual bool	set_transition_probs (SGVector< float64_t > probs)
virtual const char *	get_name () const
Public Member Functions inherited from CDistribution
	CDistribution ()
virtual	~CDistribution ()
virtual int32_t	get_num_relevant_model_parameters ()
virtual float64_t	get_log_likelihood_sample ()
virtual SGVector< float64_t >	get_log_likelihood ()
virtual float64_t	get_model_parameter (int32_t num_param)
virtual float64_t	get_derivative (int32_t num_param, int32_t num_example)
virtual float64_t	get_likelihood_example (int32_t num_example)
virtual void	set_features (CFeatures *f)
virtual CFeatures *	get_features ()
virtual void	set_pseudo_count (float64_t pseudo)
virtual float64_t	get_pseudo_count ()
Public Member Functions inherited from CSGObject
	CSGObject ()
	CSGObject (const CSGObject &orig)
virtual	~CSGObject ()
virtual bool	is_generic (EPrimitiveType *generic) const
template<class T >
void	set_generic ()
void	unset_generic ()
virtual void	print_serializable (const char *prefix="")
virtual bool	save_serializable (CSerializableFile file, const char prefix="")
virtual bool	load_serializable (CSerializableFile file, const char prefix="")
void	set_global_io (SGIO *io)
SGIO *	get_global_io ()
void	set_global_parallel (Parallel *parallel)
Parallel *	get_global_parallel ()
void	set_global_version (Version *version)
Version *	get_global_version ()
SGVector< char * >	get_modelsel_names ()
char *	get_modsel_param_descr (const char *param_name)
index_t	get_modsel_param_index (const char *param_name)

Protected Member Functions
virtual void	load_serializable_post () throw (ShogunException)

Protected Attributes
int32_t	sequence_length
int32_t	num_symbols
int32_t	num_params
float64_t *	transition_probs
float64_t *	log_transition_probs
Protected Attributes inherited from CDistribution
CFeatures *	features
float64_t	pseudo_count

Additional Inherited Members
Public Attributes inherited from CSGObject
SGIO *	io
Parallel *	parallel
Version *	version
Parameter *	m_parameters
Parameter *	m_model_selection_parameters

Constructor & Destructor Documentation

CLinearHMM ( )

default constructor

Definition at line 22 of file LinearHMM.cpp.

CLinearHMM ( CStringFeatures< uint16_t > * f )

constructor

Parameters

f	features to use

Definition at line 27 of file LinearHMM.cpp.

CLinearHMM	(	int32_t	p_num_features,
		int32_t	p_num_symbols
	)

constructor

Parameters

p_num_features	number of features
p_num_symbols	number of symbols in features

Definition at line 38 of file LinearHMM.cpp.

~CLinearHMM ( )

virtual

Definition at line 48 of file LinearHMM.cpp.

Member Function Documentation

virtual float64_t get_derivative_obsolete	(	uint16_t *	vector,
		int32_t	len,
		int32_t	pos
	)

virtual

obsolete get one example's derivative

Parameters

vector	vector
len	length
pos	position

Definition at line 131 of file LinearHMM.h.

float64_t get_likelihood_example	(	uint16_t *	vector,
		int32_t	len
	)

get one example's likelihood

Parameters

vector	the example
len	length of vector

Returns: likelihood

Definition at line 213 of file LinearHMM.cpp.

float64_t get_log_derivative	(	int32_t	num_param,
		int32_t	num_example
	)

virtual

get logarithm of one example's derivative's likelihood

Parameters

num_param	which example's param
num_example	which example

Returns: logarithm of example's derivative

Implements CDistribution.

Definition at line 223 of file LinearHMM.cpp.

virtual float64_t get_log_derivative_obsolete	(	uint16_t	obs,
		int32_t	pos
	)

virtual

obsolete get logarithm of one example's derivative's likelihood

Parameters

obs	observation
pos	position

Definition at line 119 of file LinearHMM.h.

float64_t get_log_likelihood_example	(	uint16_t *	vector,
		int32_t	len
	)

get logarithm of one example's likelihood

Parameters

vector	the example
len	length of vector

Returns: logarithm of likelihood

Definition at line 186 of file LinearHMM.cpp.

float64_t get_log_likelihood_example ( int32_t num_example )

virtual

get logarithm of one example's likelihood

Parameters

num_example which example

Returns: logarithm of example's likelihood

Implements CDistribution.

Definition at line 196 of file LinearHMM.cpp.

virtual float64_t get_log_model_parameter ( int32_t num_param )

virtual

get logarithm of given model parameter

Parameters

num_param which param

Returns: logarithm of given model parameter

Implements CDistribution.

Definition at line 173 of file LinearHMM.h.

SGVector< float64_t > get_log_transition_probs ( )

virtual

get logarithm of all transition probs

Returns: logarithm of transition probs vector

Definition at line 266 of file LinearHMM.cpp.

virtual const char* get_name ( ) const

virtual

Returns: object name

Implements CSGObject.

Definition at line 208 of file LinearHMM.h.

virtual int32_t get_num_model_parameters ( )

virtual

get number of model parameters

Returns: number of model parameters

Implements CDistribution.

Definition at line 154 of file LinearHMM.h.

virtual int32_t get_num_symbols ( )

virtual

get number of symbols in examples

Returns: number of symbols in examples

Definition at line 148 of file LinearHMM.h.

virtual float64_t get_positional_log_parameter	(	uint16_t	obs,
		int32_t	position
	)

virtual

get positional log parameter

Parameters

obs	observation
position	position

Returns: positional log parameter

Definition at line 162 of file LinearHMM.h.

virtual int32_t get_sequence_length ( )

virtual

get sequence length of each example

Returns: sequence length of each example

Definition at line 142 of file LinearHMM.h.

SGVector< float64_t > get_transition_probs ( )

virtual

get all transition probs

Returns: vector of transition probs

Definition at line 242 of file LinearHMM.cpp.

void load_serializable_post ( ) throw (ShogunException)

protectedvirtual

Can (optionally) be overridden to post-initialize some member variables which are not PARAMETER::ADD'ed. Make sure that at first the overridden method BASE_CLASS::LOAD_SERIALIZABLE_POST is called.

Exceptions

ShogunException Will be thrown if an error occurres.

Reimplemented from CSGObject.

Definition at line 290 of file LinearHMM.cpp.

bool set_log_transition_probs ( SGVector< float64_t > probs )

virtual

set logarithm of all transition probs

Parameters

probs new logarithm transition probs

Returns: if setting was successful

Definition at line 271 of file LinearHMM.cpp.

bool set_transition_probs ( SGVector< float64_t > probs )

virtual

set all transition probs

Parameters

probs new transition probs

Returns: if setting was successful

Definition at line 247 of file LinearHMM.cpp.

bool train ( CFeatures * data = NULL )

virtual

estimate LinearHMM distribution

Parameters

data	training data (parameter can be avoided if distance or kernel-based classifiers are used and distance/kernels are initialized with train data)

Returns: whether training was successful

Implements CDistribution.

Definition at line 54 of file LinearHMM.cpp.

bool train	(	const int32_t *	indizes,
		int32_t	num_indizes,
		float64_t	pseudo_count
	)

alternative train distribution

Parameters

indizes	indices
num_indizes	number of indices
pseudo_count	pseudo count

Returns: if training was successful

Definition at line 123 of file LinearHMM.cpp.

Member Data Documentation

float64_t* log_transition_probs

protected

logarithm of transition probs

Definition at line 226 of file LinearHMM.h.

int32_t num_params

protected

number of parameters

Definition at line 222 of file LinearHMM.h.

int32_t num_symbols

protected

number of symbols in examples

Definition at line 220 of file LinearHMM.h.

int32_t sequence_length

protected

examples' sequence length

Definition at line 218 of file LinearHMM.h.

float64_t* transition_probs

protected

transition probs

Definition at line 224 of file LinearHMM.h.

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

Detailed Description

Public Member Functions

Protected Member Functions

Protected Attributes

Additional Inherited Members

Constructor & Destructor Documentation

Member Function Documentation

Member Data Documentation