Open mathematicalmichael opened 5 years ago
@lcgraham do you have any thoughts on this?
@smattis do you think this is worth trying out? more features, better maintenance (active community relative to what we depend on now, which is stale).
adaptiveSampling
as mentioned in the original issue description has been fixed in #355, ignore that part, but ruling out pyDOE
as the root of the error is what led me to stumble upon pyDOE2
I am open to use pyDOE2
. We really only use the Latin hypercube sampling from pyDOE
now.
@smattis can we try to switch this out for v3?
We only use it for latin hypercube sampling and it is still working fine. If we were using more advanced features, then it may be worth the effort, but I don't see it at all as a priority.
While trying to track down the issue for #335, I found that
pyDOE
has not been maintained in several years, with the author offering an apology a little over a year ago. There are a number of issues, including lack of python3 support, and more flexible random number generation, that have languished as PRs over the years.One contributor started his own fork, which seems to be active. https://github.com/clicumu/pyDOE2
To this end, I suggest we discuss the merits of switching to
pyDOE2
in favor ofpyDOE
, or consider removing it as a dependency altogether? Still unsure if this is at all related to the build errors on https://travis-ci.org/UT-CHG/BET/jobs/536822304 but it could be, sinceadaptiveSampling
seems to be our problem.