Re: [SloppyCell-users] Inconsistent results when run in parallel mode

[Gutenkunst, R. N - (rgutenk) <rg...@em...>]

Hi Ally,

I agree, this is worrying. My big concern is if individual parameter evaluations are giving noticeably different results. The optimization routines can be sensitive to very small differences (numerical precision) in intermediate results, which is something we can't really control.

Can you send me a working example, and I can help debug?

Best,
Ryan

On May 27, 2015, at 8:23 AM, Ally Hume <a....@ed...> wrote:
> Hi,
> 
> I've been running the Optimization.fmin_lm_log_params function in parallel and sequentially and I get a different results depending on the number of nodes I use. 
> 
> I did a bit of debugging and have managed to track it down to the Calculate function in Collections.py.  If I change this part of the code so that it always runs sequentially (i.e. does all the processing locally rather than sending some off to the worker nodes) then I get consistent results no matter how many nodes I'm using.
> 
> This seems very worrying. I can probably prove data privately if you are unable to reproduce this yourselves.
> 
> Ally
> 
> Ally Hume
> Software Architect
> EPCC, The University of Edinburgh
> Tel: 0131 651 3397   Skype: ally.hume
> 
> 
> 
> 
> 
> 
> 
> 
> 
> -- 
> The University of Edinburgh is a charitable body, registered in
> Scotland, with registration number SC005336.
> 
> 
> ------------------------------------------------------------------------------
> _______________________________________________
> SloppyCell-users mailing list
> Slo...@li...
> https://lists.sourceforge.net/lists/listinfo/sloppycell-users

--
Ryan Gutenkunst
Assistant Professor
Molecular and Cellular Biology
University of Arizona
phone: (520) 626-0569, office LSS 325
http://gutengroup.mcb.arizona.edu