etl/docs/maths/pseudo_moving_average.md
2026-04-16 12:40:36 +02:00

3.5 KiB

title
pseudo_moving_average

A moving average algorithm that continuously calculates an average value from a stream of samples.
The sample size does not affect the size of the instantiated object. There is no overhead based on the number of samples as it simulates a window of N values from the current average.

There are four variants of the algorithm; two for integral values and two for floating point. Each sub-variant allows the selection of compile time or run time sample size.
The integral variant allows a compile time scaling factor to emulate fixed point arithmetic.

Integral

template <typename T, const size_t SAMPLE_SIZE, const size_t SCALING>
pseudo_moving_average;
static const size_t SAMPLE_SIZE

The number of samples averaged over.
If this value is zero, then the run time sample size specialisation is used.

static const size_t SCALING

The sample scaling factor.

pseudo_moving_average(const T initial_value)

Description
Constructs the object with the initial value for the average.


pseudo_moving_average(const T initial_value, const size_t sample_size)

Description
For runtime sample size specialisation only.
Constructs the object with the initial value for the average and the sample size.


void clear(const T initial_value)

Description
Clears the object to the initial value for the average.


void add(T new_value)

Description
Adds a new sample value.


T value() const

Description
Returns the scaled value of the average.
To unscale the returned value, use one of the rounding found in scaled_rounding.


iterator input()

Description
Returns an iterator that allows the input of new values.

Example

std::array data{ 9, 1, 8, 2, 7, 3, 6, 4, 5 };
etl::pseudo_moving_average<int, SAMPLE_SIZE, SCALING> cma(0);
std::copy(data.begin(), data.end(), cma.input());
int average = cma.value();

void set_sample_size(const size_t sample_size)

Description
For runtime sample size specialisation only.
Sets the sample size.

Floating point

template <typename T, const size_t SAMPLE_SIZE>
pseudo_moving_average;
static const size_t SAMPLE_SIZE

The number of samples averaged over.


pseudo_moving_average(const T initial_value)

Description
Constructs the object with the initial value for the average.


pseudo_moving_average(const T initial_value, const size_t sample_size)

Description
For runtime sample size specialisation only.
Constructs the object with the initial value for the average and the sample size.


void clear(const T initial_value)

Description
Clears the object to the initial value for the average.


void add(T new_value)

Description
Adds a new sample value.


T value() const

Description
Returns the average value.


void set_sample_size(const size_t sample_size)

Description
For runtime sample size specialisation only.
Sets the sample size.

How It Works

If the current moving average is 5, then an equivalent sequence of samples (for a sample size of 9), that gives the same average, would be 5, 5, 5, 5, 5, 5, 5, 5, 5

This means, to find the average when adding a new sample to a moving average that has a current value of 5, all we need to do is multiply the current average by the sample size (9), add the new sample, and divide by the sample size + 1 (10).