Closed djmce closed 4 years ago
The admixture coefficients are initialized with random values, therefore, the tool may not produce the exact same results. There is a variable 'RANDOM_START' that can be set to 0 to change this behavior.
Excellent, thanks @vibansal
I cloned the repo, found this variable in ancestry-gll.c
, set it to 0, ran make all
and tested it; confirmed via md5sum that identical outputs were obtained.
FYI, to get repeatable behaviour I also had to fix the seed to srand48()
on line 416 of ancestry-gll.c
Thanks for this update.
Does this tool involve some kind of random seeding or do you expect it to produce identical results each time it is run with the same input data?