scipy.optimize: SHGO - local minimization parameters

Hi everyone,

I have a question regarding the local minimization in SHGO from the scipy.optimize module. 

We are trying to apply a local Nelder-Mead optimization to the minimizer list of all vertex points from a sobol sequence. We would like the local minimization to terminate when either a maximum number of steps is reached or when the difference between iterations is sufficiently small. However, we are struggling to implement this. 

We are using the lmfit interface to scipy. We did the following:

minner = Minimizer(func, params, fcn_args=arg),sampling_method='sobol',n=600,minimizer_kwargs={'method':'Nelder-Mead','options':{'disp':'True','maxiter':10000}})
result = minner.minimize(method='shgo')

minner.minimize will pass all the specified options to lmfit. However 'maxiter' specified to Nelder-Mead was filtered out by the '_restrict_to_keys' function. So our question is, why is maxiter not included in the accepted keys? This seems to us the easiest way to tune the local minimization procedure. We have also discussed this at the lmfit mailing list (https://groups.google.com/g/lmfit-py/c/Fm43gjFoTCI). What would be the easiest solution for what we are trying to achieve? 


Kind regards,

Olivier
_______________________________________________
SciPy-User mailing list -- [email protected]
To unsubscribe send an email to [email protected]
https://mail.python.org/mailman3/lists/scipy-user.python.org/
Member address: [email protected]