search

simetenn / uncertainpy

Uncertainpy: a Python toolbox for uncertainty quantification and sensitivity analysis, tailored towards computational neuroscience.
http://uncertainpy.readthedocs.io
GNU General Public License v3.0
220 stars 50 forks source link

Calculation stalls at 'Calculating Statistics from PCE' #59

Open mckinnonm opened 2 years ago

mckinnonm commented 2 years ago

Hello,

I am trying to conduct PCE for uncertainty quantification for a model that produces time series data. The output is 1200 data points. I am doing UQ with 11 uncertain input parameters, and so PCE requires 2732 function calls. I have taken advantage of multiprocessing and I am currently running this job utilizing all the cores of a 48 core machine. This makes the sampling relatively fast (about 5.5 hours for all the function calls) and 'Calculating PC for each feature' requires about 3 minutes.

When it comes to 'Calculating statistics from PCE' the computation appears to stall. I have not dug into the code to see how many and what operations are required during this step, so maybe this just requires a lot of time to compute, but my machine has been sitting idle for over 12 hours with this message. Is this a reasonable amount of time for this calculation?

If this is actually indicative of an error, I would recommend including a timeout and error message into the code.

Thanks.

gdmiron commented 2 years ago

Hello,

I also had a similar experience. The "Calculating statistics from PCE" step takes aprox. 30s for 4 parameters, 3 minutes for 5 and around 1h for 6. I have not tried with more parameters. This step is not parallelized and I used a 3Ghz CPU. I wonder if this part of the code cannot be made more efficient as It seems to need a lot of time for > 6 parameters.

tfriedel commented 2 years ago

I experienced the same problem.