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 am using pymoo to find the optimal parameters for a datamatchin problem. In other words, I have a black box simulator getting parameters X as input, and returning values Y_pred as output, which need to match some ground truth experimental values Y. I cast this as an optimization problem, where
results['F'] = [np.mean(Y-Y_pred**2)]
The black box simulator is resource-intensive, so I use parallel computing. For various reasons, sometimes a simulation, ran with subprocess may time out even if it wasn't actually run, or if the run wasn't completed. In case of timeout, I'm able to catch the sp.TimeoutExpired exception.
Is it possible to tell pymoo to skip an individual?
If not, is it possible to retry the same individual once?
I don't want to assign the run a large penalty, e.g., results['F'] = 99999, because I'm not sure if the candidate was bad or not. In some cases, it may be possible that the simulation didn't run at all.
The text was updated successfully, but these errors were encountered:
AndreaPi
changed the title
Relaunch a simulation if it failed
Relaunch a simulation if it didn't run
Apr 17, 2024
I am using
pymoo
to find the optimal parameters for a datamatchin problem. In other words, I have a black box simulator getting parametersX
as input, and returning valuesY_pred
as output, which need to match some ground truth experimental valuesY
. I cast this as an optimization problem, whereresults['F'] = [np.mean(Y-Y_pred**2)]
The black box simulator is resource-intensive, so I use parallel computing. For various reasons, sometimes a simulation, ran with
subprocess
may time out even if it wasn't actually run, or if the run wasn't completed. In case of timeout, I'm able to catch thesp.TimeoutExpired
exception.pymoo
to skip an individual?I don't want to assign the run a large penalty, e.g.,
results['F'] = 99999
, because I'm not sure if the candidate was bad or not. In some cases, it may be possible that the simulation didn't run at all.The text was updated successfully, but these errors were encountered: