Testing for each algorithm could be done statistically by generating a set of initial baseline data and comparing later runs to that. That would allow much more flexibility in algorithm capabilities and more precise testing. Downsides are that a baseline dataset is needed and the testing would take longer.
Testing for each algorithm could be done statistically by generating a set of initial baseline data and comparing later runs to that. That would allow much more flexibility in algorithm capabilities and more precise testing. Downsides are that a baseline dataset is needed and the testing would take longer.