EdgeStats 0.1.0

EdgeStats

Fast statistics for edge computing.

EdgeStats is a library intended for computing statistics with sets that are subject to constant change. This is useful when computing the statistics for the last N samples of a data stream. This library greatly improves the performance of computing statistics on an edge connected device.

Supported Calulations

Though currently small, the number of calculations provided by this library will grow.

• Mean
• Sample Variance
• Sample Standard Deviation

Utility

This library provideds the following functionality:

• Adding/removing a sample from a set with O(1) time, while occupying O(1) memory.
• A sliding window class that allows for O(1) maintenance of a fixed sample size set containg the N most recently added samples. This has a memory complexity of O(N).

Performance

The bellow benchmarks were performed using BenchmarkDotNet and by taking the standard deviation and mean of a set. Increasing sample means the calculations were performed for every new sample added to the set up to SAMPLE_SIZE. Jumping sample means the calculations were performed on a set containing SAMPLE_SIZE samples, then again on a new set containing the old set plus another SAMPLE_SIZE samples. Fixed sample means the set remained unchanged and the calculations were performed only once.

Usage

Increasing sample

IncrementalStandardDeviation stats = new IncrementalStandardDeviation();

stats.AddSample(1.5);
stats.AddSample(7.6);
stats.AddSample(5.3);

// 2.515287
double standardDeviation = stats.StandardDeviation;

Decreasing sample

DecrementalStandardDeviation stats = new DecrementalStandardDeviation();

stats.AddSample(1.5);
stats.AddSample(7.6);
stats.AddSample(5.3);

// 2.515287
double standardDeviation = stats.StandardDeviation;

stats.RemoveSample(7.6);

// 1.9
standardDeviation = stats.StandardDeviation;

Sliding Window

const int WINDOW_SIZE = 10;
DecrementalStandardDeviation stats = new DecrementalStandardDeviation();
SlidingWindowStatistics slidingWindowStats = new SlidingWindowStatistics(
    stats, WINDOW_SIZE);

slidingWindowStats.AddSample(1);
...
slidingWindowStats.AddSample(10);

// 2.872281
double standardDeviation = stats.StandardDeviation;

slidingWindowStats.AddSample(15);

// 3.645545
standardDeviation = stats.StandardDeviation;

License

MIT