Closed LiuCanidk closed 4 months ago
Hi @LiuCanidk, the output that is shown above corresponds to standard CoGAPS, in the case of distributed run using nSets it should report the distributed params before the run, something like this:
-- Distributed CoGAPS Parameters --
nSets 6
cut 5
minNS 3
maxNS 9
Closing as solved, please reopen if needed.
I tried to run CoGAPS in a relatively large single cell dataset (35412 cells * 50000+ genes)
And the minumum running time for me (nPatterns, i.e., k=5, 6, 7, 8, 9, 10) was unacceptable, with "sparseOptimization=True, nSets=20". As shown below, only k=5 needs 2600+h, more than 100 days! And k=11 needs even more, ~4000h!
Did I miss something that can speed up parallelization? Or the would pyCoGAPS be much faster? (I notice in the Nature Protocol manuscript, pyCoGAPS just has a slight increase in speed performance)
Any suggestions would be greatly appreciated!