hoffmangroup / segway

Application for semi-automated genomic annotation.
http://segway.hoffmanlab.org/
GNU General Public License v2.0
13 stars 7 forks source link

Better random number generation per instance #85

Open EricR86 opened 8 years ago

EricR86 commented 8 years ago

Original report (BitBucket issue) by Eric Roberts (Bitbucket: ericr86, GitHub: ericr86).


Currently Segway has a distinct disadvantage that for that a seed N and N+1, the seeds will share same the same seeded initial parameters from N+1 onwards where instances are > 1. Other current methods seem to be error-prone or not entirely correct.

See http://stats.stackexchange.com/questions/233061/best-way-to-seed-n-independent-random-number-generators-from-1-value

Future implementations may want to use the randomstate pypi module: http://bashtage.github.io/ng-numpy-randomstate/doc/build/html/parallel.html#independent-streams Specifically it has an in-development feature of a different PRNG then current one numpy uses where you can seed and set an ID independent RNG streams.