Perturbation for the angular sampling in relion_refine/relion_refine_mpi is initialized before random number seed is set

Open bforsbe opened this issue 9 years ago • 0 comments

Originally reported by: Andreas Schenk (Bitbucket: andschenk, GitHub: andschenk)

Hi,

I encountered an issue where reruns of the exact same job with identical random seed would give different results, due to some particles being aligned differently. I narrowed down the problem to the perturbation for the angular sampling. It is initialized (using the random number generator) before the random seed is set. If I disable the perturbation (--perturb 0) or if I move the random seed initialization before the perturbation initialization (see attached patch), all reruns give the same results.

Bitbucket: https://bitbucket.org/tcblab/relion2-beta/issue/138

Oct 25 '16 08:10 bforsbe