Adaptive Stopping Rule for Performance Measurements

What is it about?

This paper shows an automated way to figure out how many times a performance benchmark needs to be repeated to obtain statistically representative results. The automated method tries to figure out how the performance data are distributed, and based on the distribution, what is the best stopping rule to apply.

Featured Image

Why is it important?

Measuring the right number of samples is important. Too few, and the results are unrepresentative or misleading; too many, and precious time and computing are wasted.