Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: Mixing reproducible and non-reproducible code



Nick MacLaren wrote, on a required reproducibility option:
>    We are also forbidding some classes of algorithm entirely, such
>    as parallel Monte-Carlo integration, which is actually the best
>    way of calculating some functions!

Not forbidden, but one of the things to be avoided by the programmer
who chooses to exercise the reproducibility option.  This is why I keep
repeating that reproducibility requires the cooperation of the programmer,
and what we want is for environments to provide the means for this in a
published manner.

The problem is that the class of things that impede reproducibility is
quite large, and if it is too large the process becomes meaningless.

The counter-problem is that if nothing is provided, it will be difficult
to consistently check even trivial things one might want to do to study
certain limited aspects of a program.

Michel.
---Sent: 2012-12-05 13:24:05 UTC