bdm::PF Class Reference

Trivial particle filter with proposal density equal to parameter evolution model. More...

#include <particles.h>

List of all members.

Public Member Functions

void set_options (const string &opt)

virtual void bayes_gensmp ()

bayes I - generate samples and add their weights to lls

virtual void bayes_weights ()

bayes II - compute weights of the

virtual bool do_resampling ()

important part of particle filtering - decide if it is time to perform resampling

void bayes (const vec &dt)

Incremental Bayes rule.

vec & __w ()

access function

vec & _lls ()

access function

RESAMPLING_METHOD _resmethod () const

const eEmp & posterior () const

access function

void from_setting (const Setting &set)

void resmethod_from_set (const Setting &set)

auxiliary function reading parameter 'resmethod' from configuration file

void prior_from_set (const Setting &set)

load prior information from set and set internal structures accordingly

void validate ()

This method TODO.

void resample (ivec &ind)

resample posterior density (from outside - see MPF)

virtual string to_string ()

This method returns a basic info about the current instance.

virtual void to_setting (Setting &set) const

This method save all the instance properties into the Setting structure.

Constructors

PF ()

void set_parameters (int n0, double res_th0=0.5, RESAMPLING_METHOD rm=SYSTEMATIC)

void set_model (shared_ptr< mpdf > par0, shared_ptr< mpdf > obs0)

void set_statistics (const vec w0, const epdf &epdf0)

void set_statistics (const eEmp &epdf0)

Constructors

virtual BM * _copy_ () const

Mathematical operations

virtual void bayesB (const mat &Dt)

Batch Bayes rule (columns of Dt are observations).

virtual double logpred (const vec &dt) const

vec logpred_m (const mat &dt) const

Matrix version of logpred.

virtual epdf * epredictor () const

Constructs a predictive density $f(d_{t+1} |d_{t}, \ldots d_{0})$ .

virtual mpdf * predictor () const

Constructs conditional density of 1-step ahead predictor $f(d_{t+1} |d_{t+h-1}, \ldots d_{t})$ .

Access to attributes

const RV & _drv () const

void set_drv (const RV &rv)

void set_rv (const RV &rv)

double _ll () const

void set_evalll (bool evl0)

Protected Attributes

int n

number of particles;

eEmp est

posterior density

vec & _w

pointer into eEmp

Array< vec > & _samples

pointer into eEmp

shared_ptr< mpdf > par

Parameter evolution model.

shared_ptr< mpdf > obs

Observation model.

vec lls

internal structure storing loglikelihood of predictions

RESAMPLING_METHOD resmethod

which resampling method will be used

double res_threshold

RV drv

Random variable of the data (optional).

double ll

Logarithm of marginalized data likelihood.

bool evalll

If true, the filter will compute likelihood of the data record and store it in ll . Set to false if you want to save computational time.

Options

bool opt_L_smp

Log all samples.

bool opt_L_wei

Log all samples.

Extension to conditional BM

This extension is useful e.g. in Marginalized Particle Filter (bdm::MPF). Alternatively, it can be used for automated connection to DS when the condition is observed

const RV & _rvc () const

access function

virtual void condition (const vec &val)

Substitute val for rvc.

RV rvc

Name of extension variable.

Logging of results

virtual void log_add (logger &L, const string &name="")

Add all logged variables to a logger.

virtual void logit (logger &L)

ivec LIDs

IDs of storages in loggers 4:[1=mean,2=lb,3=ub,4=ll].

ivec LFlags

Flags for logging - same size as LIDs, each entry correspond to the same in LIDs.

Detailed Description

Trivial particle filter with proposal density equal to parameter evolution model.

Posterior density is represented by a weighted empirical density (eEmp ).

Member Function Documentation

virtual BM* bdm::BM::_copy_ ( ) const [inline, virtual, inherited]

Copy function required in vectors, Arrays of BM etc. Have to be DELETED manually! Prototype:

 BM* _copy_() const {return new BM(*this);}

Reimplemented in bdm::ARX, bdm::ARXfrg, bdm::KalmanCh, bdm::EKFCh, and bdm::BMEF.

void bdm::PF::bayes ( const vec & dt ) [virtual]

Incremental Bayes rule.

Parameters:

dt

vector of input data

Implements bdm::BM.

References _samples, bayes_gensmp(), bayes_weights(), do_resampling(), est, lls, n, obs, bdm::eEmp::resample(), and resmethod.

void bdm::PF::from_setting ( const Setting & set ) [inline, virtual]

configuration structure for basic PF

        parameter_pdf   = mpdf_class;         // parameter evolution pdf
        observation_pdf = mpdf_class;         // observation pdf
        prior           = epdf_class;         // prior probability density
        --- optional ---
        n               = 10;                 // number of particles
        resmethod       = 'systematic', or 'multinomial', or 'stratified'
                                                                                  // resampling method
        res_threshold   = 0.5;                // resample when active particles drop below 50%

Reimplemented from bdm::BM.

References bdm::RV::add(), obs, par, prior_from_set(), and resmethod_from_set().

virtual double bdm::BM::logpred ( const vec & dt ) const [inline, virtual, inherited]

Evaluates predictive log-likelihood of the given data record I.e. marginal likelihood of the data with the posterior integrated out.

Reimplemented in bdm::ARX, bdm::MixEF, and bdm::multiBM.

References bdm_error.

Referenced by bdm::BM::logpred_m().

void bdm::PF::set_options ( const string & opt ) [inline, virtual]

Set posterior density by sampling from epdf0 Extends original BM::set_options by two more options:

logweights - meaning that all weightes will be logged
logsamples - all samples will be also logged

Reimplemented from bdm::BM.

References opt_L_smp, and opt_L_wei.

Member Data Documentation

double bdm::PF::res_threshold [protected]

resampling threshold; in this case its meaning is minimum ratio of active particles For example, for 0.5 resampling is performed when the numebr of active aprticles drops belo 50%.

Referenced by do_resampling(), and resmethod_from_set().

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

particles.h
particles.cpp


Public Member Functions
void	set_options (const string &opt)
virtual void	bayes_gensmp ()
	bayes I - generate samples and add their weights to lls
virtual void	bayes_weights ()
	bayes II - compute weights of the
virtual bool	do_resampling ()
	important part of particle filtering - decide if it is time to perform resampling
void	bayes (const vec &dt)
	Incremental Bayes rule.
vec &	__w ()
	access function
vec &	_lls ()
	access function
RESAMPLING_METHOD	_resmethod () const
const eEmp &	posterior () const
	access function
void	from_setting (const Setting &set)
void	resmethod_from_set (const Setting &set)
	auxiliary function reading parameter 'resmethod' from configuration file
void	prior_from_set (const Setting &set)
	load prior information from set and set internal structures accordingly
void	validate ()
	This method TODO.
void	resample (ivec &ind)
	resample posterior density (from outside - see MPF)
virtual string	to_string ()
	This method returns a basic info about the current instance.
virtual void	to_setting (Setting &set) const
	This method save all the instance properties into the Setting structure.
Constructors
	PF ()
void	set_parameters (int n0, double res_th0=0.5, RESAMPLING_METHOD rm=SYSTEMATIC)
void	set_model (shared_ptr< mpdf > par0, shared_ptr< mpdf > obs0)
void	set_statistics (const vec w0, const epdf &epdf0)
void	set_statistics (const eEmp &epdf0)
Constructors
virtual BM *	_copy_ () const
Mathematical operations
virtual void	bayesB (const mat &Dt)
	Batch Bayes rule (columns of Dt are observations).
virtual double	logpred (const vec &dt) const
vec	logpred_m (const mat &dt) const
	Matrix version of logpred.
virtual epdf *	epredictor () const
	Constructs a predictive density $f(d_{t+1} \|d_{t}, \ldots d_{0})$ .
virtual mpdf *	predictor () const
	Constructs conditional density of 1-step ahead predictor $f(d_{t+1} \|d_{t+h-1}, \ldots d_{t})$ .
Access to attributes
const RV &	_drv () const
void	set_drv (const RV &rv)
void	set_rv (const RV &rv)
double	_ll () const
void	set_evalll (bool evl0)
Protected Attributes
int	n
	number of particles;
eEmp	est
	posterior density
vec &	_w
	pointer into `eEmp`
Array< vec > &	_samples
	pointer into `eEmp`
shared_ptr< mpdf >	par
	Parameter evolution model.
shared_ptr< mpdf >	obs
	Observation model.
vec	lls
	internal structure storing loglikelihood of predictions
RESAMPLING_METHOD	resmethod
	which resampling method will be used
double	res_threshold
RV	drv
	Random variable of the data (optional).
double	ll
	Logarithm of marginalized data likelihood.
bool	evalll
	If true, the filter will compute likelihood of the data record and store it in `ll` . Set to false if you want to save computational time.
Options
bool	opt_L_smp
	Log all samples.
bool	opt_L_wei
	Log all samples.
Extension to conditional BM
This extension is useful e.g. in Marginalized Particle Filter (bdm::MPF). Alternatively, it can be used for automated connection to DS when the condition is observed
const RV &	_rvc () const
	access function
virtual void	condition (const vec &val)
	Substitute `val` for `rvc`.
RV	rvc
	Name of extension variable.
Logging of results
virtual void	log_add (logger &L, const string &name="")
	Add all logged variables to a logger.
virtual void	logit (logger &L)
ivec	LIDs
	IDs of storages in loggers 4:[1=mean,2=lb,3=ub,4=ll].
ivec	LFlags
	Flags for logging - same size as LIDs, each entry correspond to the same in LIDs.