Infuse-IoT running statistics APIs

Running statistics API . More...

Data Structures
struct	statistics_state

Functions
static void	statistics_reset (struct statistics_state *state)
	Reset statistics object.
void	statistics_update (struct statistics_state *state, int32_t value)
	Update the statistics object with a new sample.
static float	statistics_mean (const struct statistics_state *state)
	Compute the mean of the statistics object.
static float	statistics_variance (const struct statistics_state *state)
	Compute the variance of the statistics object.
static int32_t	statistics_mean_rough (const struct statistics_state *state)
	Compute the rough mean of the statistics object.
static uint64_t	statistics_variance_rough (const struct statistics_state *state)
	Compute the rough variance of the statistics object.

Detailed Description

Running statistics API .

This API is a C implementation of the following paper, which is itself inspired by the well known Welford's algorithm: https://sfat.massey.ac.nz/research/centres/crisp/pdfs/2013_IVCNZ_214.pdf

Function Documentation

statistics_mean()

static float statistics_mean ( const struct statistics_state * state )

inlinestatic

#include <infuse/math/statistics.h>

Compute the mean of the statistics object.

Parameters

state

Statistics state object

Returns

float Mean of samples

statistics_mean_rough()

static int32_t statistics_mean_rough ( const struct statistics_state * state )

inlinestatic

#include <infuse/math/statistics.h>

Compute the rough mean of the statistics object.

The computed value is "rough" in the sense that it does not attempt to round to the nearest whole number, and merely takes the integer portion.

Tests validate that this value is within 1.0f of the value returned by statistics_mean.

Parameters

state

Statistics state object

Returns

uint32_t Approximate mean of samples

statistics_reset()

static void statistics_reset ( struct statistics_state * state )

inlinestatic

#include <infuse/math/statistics.h>

Reset statistics object.

Parameters

state

Statistics state object to reset

statistics_update()

void statistics_update	(	struct statistics_state *	state,
		int32_t	value )

#include <infuse/math/statistics.h>

Update the statistics object with a new sample.

Parameters

state	Statistics state object
value	New value to feed into state

statistics_variance()

static float statistics_variance ( const struct statistics_state * state )

inlinestatic

#include <infuse/math/statistics.h>

Compute the variance of the statistics object.

Parameters

state

Statistics state object

Returns

float Variance of samples

statistics_variance_rough()

static uint64_t statistics_variance_rough ( const struct statistics_state * state )

inlinestatic

#include <infuse/math/statistics.h>

Compute the rough variance of the statistics object.

The computed value is "rough" in the sense that it does not attempt to round to the nearest whole number, and merely takes the integer portion. It also does not take into account the difference between state->m and the true mean, as described in the paper.

Tests validate that this value is within 1.0f or 2% of the value returned by statistics_variance, whichever is less accurate.

Parameters

state

Statistics state object

Returns

uint64_t Approximate variance of samples