In file ../include/EST_simplestats.h:

class EST_DiscreteProbDistribution

A class for representing probability distributions for a set of discrete values.

Public Methods

	EST_DiscreteProbDistribution (const EST_DiscreteProbDistribution &b) Create with copying from an existing distribution
	EST_DiscreteProbDistribution (const EST_TList<EST_String> &vocab) Create with given vocabulary
	EST_DiscreteProbDistribution (const EST_Discrete* d) Create using given EST_Discrete class as the vocabulary
	EST_DiscreteProbDistribution (const EST_Discrete* d, const double n_samples, const EST_DVector &counts) Create using given EST_Discrete class as vocabulary plus given counts
	~EST_DiscreteProbDistribution () Destructor function
void	copy (const EST_DiscreteProbDistribution &b) Copy all data from another DPD to this
void	clear (void) Reset, clearing all counts and vocabulary
bool	init (const EST_StrList &vocab) Initialise using given vocabulary
void	init (const EST_Discrete* d) Initialise using given EST_Distcrete as vocabulary
void	init () Initialise
double	samples (void) const Total number of example found
void	cumulate (const EST_String &s, double count=1) Add this observation, may specify number of occurences
void	cumulate (const int i, double count=1) Add this observation, i must be with in EST_Discrete range
const EST_String&	most_probable (double* prob = NULL) const Return the most probable member of the distribution
double	entropy (void) const Return the entropy of the distribution
double	probability (const EST_String &s) const
double	probability (const int i) const
double	frequency (const EST_String &s) const
double	frequency (const int i) const
int	item_start () const Used for iterating through members of the distribution
int	item_next (int idx) const Used for iterating through members of the distribution
int	item_end (int idx) const Used for iterating through members of the distribution
const EST_String&	item_name (int idx) const During iteration ruturns name given index
void	item_freq (int idx, EST_String &s, double &freq) const During iteration ruturns name and frequency given index
void	item_prob (int idx, EST_String &s, double &prob) const During iteration ruturns name and probability given index
inline const EST_Discrete* const	get_discrete () const Returns discrete vocabilary of distribution
void	set_frequency (const EST_String &s, double c) Sets the frequency of named item, modifes num_samples accordingly.
void	set_frequency (int i, double c) Sets the frequency of named item, modifes num_samples accordingly.
void	override_frequency (const EST_String &s, double c) Sets the frequency of named item, without modifying num_samples
void	override_frequency (int i, double c) Sets the frequency of named item, without modifying num_samples
void	set_num_samples (const double c) Sets the number of samples.

Documentation

A class for representing probability distributions for a set of discrete values.

This may be used to cummulate the probability distribution of a class of values. Values are actually help as frequencies so both frequency and probability information may be available. Note that frequencies are not integers because usings smoothing and backoff integers are too restrictive so they are actually represented as doubles.

Methods are provided to iterate over the values in a distribution, for example

       EST_DiscreteProbistribution pdf;
       for (int i=pdf.item_start(); i < pdf.item_end(); i=pdf.item_next(i))
       {
          EST_String name;
          double prob;
          item_prob(i,name,prob);
          cout << name << ": prob " << prob << endl;
       }
    

EST_DiscreteProbDistribution(const EST_DiscreteProbDistribution &b)

Create with copying from an existing distribution

EST_DiscreteProbDistribution(const EST_TList<EST_String> &vocab)

Create with given vocabulary

EST_DiscreteProbDistribution(const EST_Discrete* d)

Create using given EST_Discrete class as the vocabulary

EST_DiscreteProbDistribution(const EST_Discrete* d, const double n_samples, const EST_DVector &counts)

Create using given EST_Discrete class as vocabulary plus given counts

~EST_DiscreteProbDistribution()

Destructor function

void copy(const EST_DiscreteProbDistribution &b)

Copy all data from another DPD to this

void clear(void)

Reset, clearing all counts and vocabulary

bool init(const EST_StrList &vocab)

Initialise using given vocabulary

void init(const EST_Discrete* d)

Initialise using given EST_Distcrete as vocabulary

void init()

Initialise

double samples(void) const

Total number of example found

void cumulate(const EST_String &s, double count=1)

Add this observation, may specify number of occurences

void cumulate(const int i, double count=1)

Add this observation, i must be with in EST_Discrete range

const EST_String& most_probable(double* prob = NULL) const

Return the most probable member of the distribution

double entropy(void) const

Return the entropy of the distribution

double probability(const EST_String &s) const

double probability(const int i) const

double frequency(const EST_String &s) const

double frequency(const int i) const

int item_start() const

Used for iterating through members of the distribution

int item_next(int idx) const

Used for iterating through members of the distribution

int item_end(int idx) const

Used for iterating through members of the distribution

const EST_String& item_name(int idx) const

During iteration ruturns name given index

void item_freq(int idx, EST_String &s, double &freq) const

During iteration ruturns name and frequency given index

void item_prob(int idx, EST_String &s, double &prob) const

During iteration ruturns name and probability given index

inline const EST_Discrete* const get_discrete() const

Returns discrete vocabilary of distribution

void set_frequency(const EST_String &s, double c)

Sets the frequency of named item, modifes num_samples accordingly. This is used when smoothing frequencies.

void set_frequency(int i, double c)

Sets the frequency of named item, modifes num_samples accordingly. This is used when smoothing frequencies.

void override_frequency(const EST_String &s, double c)

Sets the frequency of named item, without modifying num_samples

void override_frequency(int i, double c)

Sets the frequency of named item, without modifying num_samples

void set_num_samples(const double c)

Sets the number of samples. Care should be taken on setting this as it will affect how probabilities are calculated.

This class has no child classes.

Friends:

ostream & operator <<(ostream &s, const EST_DiscreteProbDistribution &p)

Author:

Alan W Black (awb@cstr.ed.ac.uk): July 1996

Alphabetic index HTML hierarchy of classes or Java

This page is part of the Edinburgh Speech Tools Library documentation
Copyright University of Edinburgh 1997
Contact: speech_tools@cstr.ed.ac.uk