Reproducibility in parallel? about cola HOT 3 CLOSED

Now I use doRNG package for parallel computing. In theory if you set the same seed by set.seed(), the classifications will be the same.

PS. the objects might not be the same because the objects also contain functions and environments, but for the non-environment parts in the object, I just had a quick check and they are the same if the objects are generated from the same seed.

from cola.

You are right. When multiple cores are turned one, you cannot set a random seed for that.

Let me find a solution...

from cola.

Now I use doRNG package for parallel computing. In theory if you set the same seed by set.seed(), the classifications will be the same.

PS. the objects might not be the same because the objects also contain functions and environments, but for the non-environment parts in the object, I just had a quick check and they are the same if the objects are generated from the same seed.

I installed the latest version from github 2.5.1 and now setting a seed before run_all_consensus_partition_methods() leads to reproducible clusters when parallel computing. Thanks!

from cola.