mirror of
https://github.com/davisking/dlib.git
synced 2024-11-01 10:14:53 +08:00
Added the running_stats object
--HG-- extra : convert_revision : svn%3Afdd8eb12-d10e-0410-9acb-85c331704f74/trunk%402340
This commit is contained in:
Davis King
parent
e8537b1f7b
commit
cd9dbd575a
11
dlib/statistics.h
Normal file
11
dlib/statistics.h
Normal file
@ -0,0 +1,11 @@
|
||||
// Copyright (C) 2008 Davis E. King (davisking@users.sourceforge.net)
|
||||
// License: Boost Software License See LICENSE.txt for the full license.
|
||||
#ifndef DLIB_STATISTICs_H_
|
||||
#define DLIB_STATISTICs_H_
|
||||
|
||||
#include "statistics/statistics.h"
|
||||
|
||||
#endif // DLIB_STATISTICs_H_
|
||||
|
||||
|
||||
|
||||
120
dlib/statistics/statistics.h
Normal file
120
dlib/statistics/statistics.h
Normal file
@ -0,0 +1,120 @@
|
||||
// Copyright (C) 2008 Davis E. King (davisking@users.sourceforge.net)
|
||||
// License: Boost Software License See LICENSE.txt for the full license.
|
||||
#ifndef DLIB_STATISTICs_
|
||||
#define DLIB_STATISTICs_
|
||||
|
||||
#include "statistics_abstract.h"
|
||||
#include <limits>
|
||||
#include <cmath>
|
||||
#include "../algs.h"
|
||||
|
||||
namespace dlib
|
||||
{
|
||||
|
||||
// ----------------------------------------------------------------------------------------
|
||||
|
||||
template <
|
||||
typename T
|
||||
>
|
||||
class running_stats
|
||||
{
|
||||
public:
|
||||
|
||||
running_stats()
|
||||
{
|
||||
clear();
|
||||
|
||||
COMPILE_TIME_ASSERT ((
|
||||
is_same_type<float,T>::value ||
|
||||
is_same_type<double,T>::value ||
|
||||
is_same_type<long double,T>::value
|
||||
));
|
||||
}
|
||||
|
||||
void clear()
|
||||
{
|
||||
sum = 0;
|
||||
sum_sqr = 0;
|
||||
n = 0;
|
||||
maximum_n = std::numeric_limits<T>::max();
|
||||
}
|
||||
|
||||
void set_max_n (
|
||||
const T& val
|
||||
)
|
||||
{
|
||||
maximum_n = val;
|
||||
}
|
||||
|
||||
void add (
|
||||
const T& val
|
||||
)
|
||||
{
|
||||
const T div_n = 1/(n+1);
|
||||
const T n_div_n = n*div_n;
|
||||
|
||||
sum = n_div_n*sum + val*div_n;
|
||||
sum_sqr = n_div_n*sum_sqr + val*div_n*val;
|
||||
|
||||
if (n < maximum_n)
|
||||
++n;
|
||||
}
|
||||
|
||||
T max_n (
|
||||
) const
|
||||
{
|
||||
return max_n;
|
||||
}
|
||||
|
||||
T current_n (
|
||||
) const
|
||||
{
|
||||
return n;
|
||||
}
|
||||
|
||||
T mean (
|
||||
) const
|
||||
{
|
||||
return sum;
|
||||
}
|
||||
|
||||
T variance (
|
||||
) const
|
||||
{
|
||||
// make sure requires clause is not broken
|
||||
DLIB_ASSERT(current_n() > 1,
|
||||
"\tT running_stats::variance"
|
||||
<< "\n\tsize of queue should not be zero"
|
||||
<< "\n\tthis: " << this
|
||||
);
|
||||
|
||||
const T temp = n/(n-1);
|
||||
return temp*(sum_sqr - sum*sum);
|
||||
}
|
||||
|
||||
T scale (
|
||||
const T& val
|
||||
) const
|
||||
{
|
||||
// make sure requires clause is not broken
|
||||
DLIB_ASSERT(current_n() > 1,
|
||||
"\tT running_stats::variance"
|
||||
<< "\n\tsize of queue should not be zero"
|
||||
<< "\n\tthis: " << this
|
||||
);
|
||||
return (val-mean())/std::sqrt(variance());
|
||||
}
|
||||
|
||||
private:
|
||||
T sum;
|
||||
T sum_sqr;
|
||||
T n;
|
||||
T maximum_n;
|
||||
};
|
||||
|
||||
// ----------------------------------------------------------------------------------------
|
||||
|
||||
}
|
||||
|
||||
#endif // DLIB_STATISTICs_
|
||||
|
||||
132
dlib/statistics/statistics_abstract.h
Normal file
132
dlib/statistics/statistics_abstract.h
Normal file
@ -0,0 +1,132 @@
|
||||
// Copyright (C) 2008 Davis E. King (davisking@users.sourceforge.net)
|
||||
// License: Boost Software License See LICENSE.txt for the full license.
|
||||
#undef DLIB_STATISTICs_ABSTRACT_
|
||||
#ifdef DLIB_STATISTICs_ABSTRACT_
|
||||
|
||||
#include <limits>
|
||||
#include <cmath>
|
||||
|
||||
namespace dlib
|
||||
{
|
||||
|
||||
template <
|
||||
typename T
|
||||
>
|
||||
class running_stats
|
||||
{
|
||||
/*!
|
||||
REQUIREMENTS ON T
|
||||
- T must be a float, double, or long double type
|
||||
|
||||
INITIAL VALUE
|
||||
- max_n() == std::numeric_limits<T>::max()
|
||||
- mean() == 0
|
||||
- current_n() == 0
|
||||
|
||||
WHAT THIS OBJECT REPRESENTS
|
||||
This object represents something that can compute the running mean and
|
||||
variance of a stream of real numbers.
|
||||
|
||||
As this object accumulates more and more numbers it will be the case
|
||||
that each new number impacts the current mean and variance estimate
|
||||
less and less. This may be what you want. But it might not be.
|
||||
|
||||
For example, your stream of numbers might be non-stationary, that is,
|
||||
the mean and variance might change over time. To enable you to use
|
||||
this object on such a stream of numbers this object provides the
|
||||
ability to set a "max_n." The meaning of the max_n() parameter
|
||||
is that after max_n() samples have been seen each new sample will
|
||||
have the same impact on the mean and variance estimates from then on.
|
||||
|
||||
So if you have a highly non-stationary stream of data you might
|
||||
set the max_n to a small value while if you have a very stationary
|
||||
stream you might set it to a very large value.
|
||||
!*/
|
||||
public:
|
||||
|
||||
running_stats(
|
||||
);
|
||||
/*!
|
||||
ensures
|
||||
- this object is properly initialized
|
||||
!*/
|
||||
|
||||
void clear(
|
||||
);
|
||||
/*!
|
||||
ensures
|
||||
- this object has its initial value
|
||||
- clears all memory of any previous data points
|
||||
!*/
|
||||
|
||||
void set_max_n (
|
||||
const T& val
|
||||
);
|
||||
/*!
|
||||
ensures
|
||||
- #max_n() == val
|
||||
!*/
|
||||
|
||||
T max_n (
|
||||
) const;
|
||||
/*!
|
||||
ensures
|
||||
- returns the max value that current_n() is allowed to take on
|
||||
!*/
|
||||
|
||||
T current_n (
|
||||
) const;
|
||||
/*!
|
||||
ensures
|
||||
- returns the number of points given to this object so far or
|
||||
max_n(), whichever is smallest.
|
||||
!*/
|
||||
|
||||
void add (
|
||||
const T& val
|
||||
);
|
||||
/*!
|
||||
ensures
|
||||
- updates the mean and variance stored in this object so that
|
||||
the new value is factored into them
|
||||
- #mean() == mean()*current_n()/(current_n()+1) + val/(current_n()+1)
|
||||
- #variance() == the updated variance that takes this new value into account
|
||||
- if (current_n() < max_n()) then
|
||||
- #current_n() == current_n() + 1
|
||||
- else
|
||||
- #current_n() == current_n()
|
||||
!*/
|
||||
|
||||
T mean (
|
||||
) const;
|
||||
/*!
|
||||
ensures
|
||||
- returns the mean of all the values presented to this object
|
||||
so far.
|
||||
!*/
|
||||
|
||||
T variance (
|
||||
) const;
|
||||
/*!
|
||||
requires
|
||||
- current_n() > 1
|
||||
ensures
|
||||
- returns the variance of all the values presented to this
|
||||
object so far.
|
||||
!*/
|
||||
|
||||
T scale (
|
||||
const T& val
|
||||
) const;
|
||||
/*!
|
||||
requires
|
||||
- current_n() > 1
|
||||
ensures
|
||||
- return (val-mean())/std::sqrt(variance());
|
||||
!*/
|
||||
};
|
||||
|
||||
}
|
||||
|
||||
#endif // DLIB_STATISTICs_ABSTRACT_
|
||||
|
||||
Loading…
Reference in New Issue
Block a user