I found out after painful debugging that dlib's find_global_min fails with a segmentation fault when the observed values contain NaN.
What do you think about including a NaN checking procedure to optimization algorithms?
If you like the idea I'll work on it and commit to dlib.
It certainly shouldn't segfault. So yeah, if you want to submit a PR that makes it throw dlib::error or puts a DLIB_CASSERT on this then that would be a good addition.
I found out after painful debugging that dlib's
find_global_min
fails with a segmentation fault when the observed values containNaN
. What do you think about including aNaN
checking procedure to optimization algorithms? If you like the idea I'll work on it and commit to dlib.