You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I recently realized that getPropertySamples is very slow (43s for 2k sample iterations) in the conversionReaction example.
The largest portion (29s) of time goes into the 21 calls of plotPropertyUncertainty which updates the legend and histogram bars each time. Indeed, it updates every 100 MCMC iterations as coded in line 119 of getPropertySamples:
if (mod(j,100) == 0) || (j == length(properties.S.logPost))
I suggest to change it to 10%, 20%, ... of the iterations instead or removing the sequential drawing entirely.
The text was updated successfully, but these errors were encountered:
Ok, I changed the code to a percentage heuristic for now and gonna merge it into development in near future. An adaptive strategy sounds good. I'd suggest we implement it together with the modularization / expansion of the visularization tools as mentioned in issue #49
I recently realized that getPropertySamples is very slow (43s for 2k sample iterations) in the conversionReaction example.
The largest portion (29s) of time goes into the 21 calls of plotPropertyUncertainty which updates the legend and histogram bars each time. Indeed, it updates every 100 MCMC iterations as coded in line 119 of getPropertySamples:
I suggest to change it to 10%, 20%, ... of the iterations instead or removing the sequential drawing entirely.
The text was updated successfully, but these errors were encountered: