-
Notifications
You must be signed in to change notification settings - Fork 125
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
using iterative minimizer routine to improve getting stuck in a local minimum #277
Comments
If you run with MPI it should try multiple random points are report the lowest, with a warning if it doesn't look well converged. You can of course also play with the accuracy parameters |
Thanks for your reply, Antony. My tests are always with MPI (np=6) in a single minimize round, and it can still finish successfully without any warning, although the result is definitely not good enough (since I found a better one using the iterative routine). |
I notice "minimize" can easily get stuck in a local minimum if the theory model is complicated even with the covmat provided by MCMC chains.
One piece of evidence is that if I run "minimize" again with exactly the same setting (rhoend=0.01), but the starting point from the last minimum, it can return a better result (sometimes chi2 improvement can be more than 1).
I found one way to improve it is to use the iterative minimizer routine: repeat "minimize" by setting the starting point as the last minimum until the chi2 difference between the current and last round is less than a certain threshold, e.g. 0.1. This iterative routine can significantly improve the result, sometimes by Dchi2~3 (e.g. in Class Early Dark Energy scalar field model).
The text was updated successfully, but these errors were encountered: