Generic statistical utilities. More...

Macros
#define	EXP_A (1048576.0 / M_LN2)

#define	EXP_C 60801

#define	M_LN2 0.69314718055994530942

Functions
void	drwnInitializeRand ()
	initialize the standard C library random number generator with a time-of-day seed

template<typename T >
T	drwn::minElem (const vector< T > &v)
	returns the minimum element in a vector of objects

template<typename T >
T	drwn::maxElem (const vector< T > &v)
	returns the maximum element in a vector of objects

template<typename T >
T	drwn::mean (const vector< T > &v)
	returns the mean of all elements in a vector of objects

template<typename T >
T	drwn::median (const vector< T > &v)
	returns the median element in a vector of objects

template<typename T >
T	drwn::destructive_median (vector< T > &w)
	returns the median element in a vector of objects (but may modify the vector's contents)

template<typename T >
T	drwn::mode (const vector< T > &v)
	returns the most frequent element in a vector of objects

template<typename T >
T	drwn::variance (const vector< T > &v)
	returns the variance (second moment about the mean) of all elements in a vector of objects

template<typename T >
T	drwn::stdev (const vector< T > &v)
	returns the standard deviation of all elements in a vector of objects

template<typename T >
int	drwn::argmin (const vector< T > &v)
	returns the index of the smallest element in a vector of objects

int	drwn::argmin (const VectorXd &v)
	returns the index of the smallest element in a vector of objects

template<typename T >
vector< int >	drwn::argmins (const vector< vector< T > > &v)
	returns the index for the smallest element in each of vector of vector of objects

template<typename T >
int	drwn::argmax (const vector< T > &v)
	returns the index of the largest element in a vector of objects

int	drwn::argmax (const VectorXd &v)
	returns the index of the largest element in a vector of objects

template<typename T >
vector< int >	drwn::argmaxs (const vector< vector< T > > &v)
	returns the index for the largest element in each of vector of vector of objects

int	drwn::argrand (const vector< double > &v)
	returns the index for a random element sampled in proportion to the size of the element from a vector of positive entries

int	drwn::argrand (const VectorXd &v)
	returns the index for a random element sampled in proportion to the size of the element from a vector of positive entries

template<typename T >
T	drwn::excessKurtosis (const vector< T > &v)
	returns the kurtosis for a vector of objects

template<typename T >
vector< float >	drwn::percentiles (const vector< T > &v)

template<typename T >
pair< T, T >	drwn::range (const vector< T > &v)
	returns the minimum and maximum values in a vector of objects

template<typename T >
pair< T, T >	drwn::range (const vector< vector< T > > &v)
	returns the minimum and maximum values in a vector of vector of objects

template<typename T >
vector< T >	drwn::extractSubVector (const vector< T > &v, const vector< int > &indx)
	select an ordered subvector from a vector

template<typename T >
vector< T >	drwn::removeOutliers (const vector< T > &v, const vector< double > &scores, int keepSize)
	removes (v.size() - keepSize)/2 minimum and maximum entries

template<typename T >
set< set< T > >	drwn::powerset (const set< T > &s)
	generate powerset of a set

int	drwn::roundUp (int n, int d)
	rounds (away from zero) to nearest discretization

bool	drwn::containsInvalidEntries (const vector< double > &v)
	returns true if the vector contains NaN or Inf values

double	drwn::logistic (const vector< double > &theta, const vector< double > &data)
	logistic function $y = \frac{1}{1 + \exp\left\{- \theta^T x\right\}}$

double	drwn::logistic (const double theta, const double data, int n)
	logistic function $y = \frac{1}{1 + \exp\left\{- \theta^T x\right\}}$

double	drwn::entropy (const std::vector< double > &p)
	computes the entropy of a possibly unnormalized distribution

double	drwn::entropy (const std::vector< int > &counts)
	computes the entropy of a frequency histogram

double	drwn::gini (const std::vector< double > &p)
	computes the gini impurity of a possibly unnormalized distribution

double	drwn::gini (const std::vector< int > &p)
	computes the gini impurity of a frequency histogram

double	drwn::expAndNormalize (std::vector< double > &v)
	exponentiates and normalizes a vector in-place; returns log of the normalization constant

double	drwn::expAndNormalize (VectorXd &v)
	exponentiates and normalizes a vector in-place; returns log of the normalization constant

double	drwn::fastexp (double x)
	fast exponentiation

vector< int >	drwn::randomPermutation (int n)
	compute a random permutation of the numbers [0..n-1]

template<typename T >
void	drwn::shuffle (vector< T > &v)
	randomly permutes the entries of a vector inline

template<typename T >
vector< T >	drwn::subSample (const vector< T > &v, size_t n)
	extract a subsample from a vector of size `n`

vector< double >	drwn::linSpaceVector (double startValue, double endValue, unsigned n=10)
	generate a vector of linearly-spaced values from `startValue` to `endValue`

vector< double >	drwn::logSpaceVector (double startValue, double endValue, unsigned n=10)
	generate a vector of logarithmically-spaced values from `startValue` to `endValue`

void	drwn::predecessor (std::vector< int > &array, int limit)
	Computes the predecessor of a discrete vector, for example, predecessor([1 0 0], 2) produces [0 0 0]. Each position must have the same cardinality.

void	drwn::successor (std::vector< int > &array, int limit)
	Computes the successor of a discrete vector, for example, successor([1 0 0], 2) produces [0 1 0]. Each position must have the same cardinality.

void	drwn::predecessor (std::vector< int > &array, const std::vector< int > &limits)
	Computes the predecessor of a discrete vector, for example, predecessor([1 0 0], [2 2 2]) produces [0 0 0]. Each position can have different cardinality.

void	drwn::successor (std::vector< int > &array, const std::vector< int > &limits)
	Computes the successor of a discrete vector, for example, successor([1 0 0], [2 2 2]) produces [0 1 0]. Each position can have different cardinality.

double	drwn::huberFunction (double x, double m=1.0)
	huber penalty function, for $\|x\| \leq m$ and $\textrm{sgn}(x) m (2x - m)$ otherwise

double	drwn::huberDerivative (double x, double m=1.0)
	derivative of huberFunction at x

double	drwn::huberFunctionAndDerivative (double x, double *df, double m=1.0)
	huber penalty function and derivative at x

double	drwn::bhattacharyyaDistance (std::vector< double > &p, std::vector< double > &q)
	Computes the Bhattacharyya distance between two discrete probability distributions. The distributions do not need to be normalized.

double	drwn::euclideanDistanceSq (std::vector< double > &p, std::vector< double > &q)
	Computes the Euclidean norm between two discrete probability distributions. The distributions do not need to be normalized.

double	drwn::sum (const vector< double > &v)
	sum the elements in a vector

double	drwn::sum (const double *v, size_t length)
	sum the elements in a vector

double	drwn::dot (const double x, const double y, size_t length)
	dot product between elements in two vectors

double	drwn::dot (const vector< double > &x, const vector< double > &y)
	dot product between elements in two vectors

bool	drwn::eq (const double x, const double y)
	whether two numbers are equal

bool	drwn::lt (const double x, const double y)
	whether first number is less than second number

Detailed Description

Generic statistical utilities.

Function Documentation

◆ percentiles()

template<typename T >

vector< float > drwn::percentiles ( const vector< T > & v )

Todo:: can change from O(n^2) to O(n log n) by using a sorting implementation

Macros

Functions

Detailed Description

Function Documentation

◆ percentiles()