hmmm .. I agree and disagree. Yes code is often the liability, but at least in science, the code is a manifestion of your hypothesis, of your science. - Deepak Singh

As a reproducibility advocate for scientific research, I will probably negatively judge code that chooses a probability-based approach when a deterministic implementation is feasible. At some point, though, you gotta get the job done. "Good enough for government work." - Chris Lasher

Surely as long as the dataset (the exact one used) is available for comparison along with the code this is reproducible enough? Requires proper versioning and sharing of datasets of course but I doubt we would argue over the need for that :-) - Cameron Neylon

This point is also, IIRC, one of the principles in "The Pragmatic Programmer", by Hunt and Thomas. - Michael Nielsen