method selection does not work when calling minimize directly

[ghost]

Hi there,

great work. This interface really improves the handling of numpy.optimize. The issue: When passing method="nelder" to minimize, a leastsquare fit is performed. This is due to the code fragment line 498 in minimizer.py. The method argument is only parsed if it refers to a scalar function (HAS_SCALAR_MIN is true).

I suggest replacing

    if not found:
        fitter.leastsq()

by

    if not found:
        if meth in _methods:
            func = 'fitter.'+_methods[meth]+'(**fit_kws)'
            found = True
            exec(func)
        else:
            fitter.leastsq(**fit_kws)

Best regards,

Alois

[newville]

Hi Alois, thanks for the message.

I believe this is already fixed in the git repo. That is, specifying 'nelder' and the other scaler minimization procedures works for me.

[ghost]

Hey,

I checked only on the newest version in the master branch which pops up immediately on the github side (is it the newest version?). There was a commit one month ago which resulted in a deletion of a respective code block:

commit d02e8c7e1ec733ce7ed842b8b0e8b24b19049b55

Am I getting something wrong here?

Best regards,

Alois

On Wed, Oct 17, 2012 at 4:54 PM, Matt Newville notifications@github.comwrote:

Hi Alois, thanks for the message.

I believe this is already fixed in the git repo. That is, specifying 'nelder' and the other scaler minimization procedures works for me.

— Reply to this email directly or view it on GitHubhttps://github.com/newville/lmfit-py/issues/25#issuecomment-9530235.

[newville]

I added tests/fit_pvoight_NelderMead2.py This works for me from the master branch. If something isn't working for you, please send a small script that fails on the master branch. That's much better than trying to guess what is going wrong or posting partial code.

[ghost]

Hey there,

sorry to bother you again, but: I installed the master branch from the repo and tested your script. It looks like it still executes the leastsquare algorithm.

I added report_errors(myfit.params) to your script and there are error values on each parameter which are not given by the fmin algorithm.

To verify, I added

print "I'm a least square fit!"

to the leastsq function in minimize. This is the output:

$:~/phd/somebody/trunk/tools/lmfit-master/lmfit-py/tests$ python fit_pvoigt_NelderMead2.py I am a least square fit! (' Nfev = ', 50) amp_g: 21.096483 +/- 0.052981 (0.25%) initial = 10.000000 amp_l: 21.096483 +/- 0.000000 (0.00%) == 'amp_g' cen_g: 8.097970 +/- 0.002799 (0.03%) initial = 9.000000 cen_l: 8.097970 +/- 0.000000 (0.00%) == 'cen_g' frac: 0.399482 +/- 0.013247 (3.32%) initial = 0.500000 line_off: -1.126788 +/- 0.035736 (3.17%) initial = 0.000000 line_slope: 0.623733 +/- 0.001930 (0.31%) initial = 0.000000 wid_g: 1.600233 +/- 0.004338 (0.27%) initial = 1.000000 wid_l: 1.600233 +/- 0.000000 (0.00%) == 'wid_g' Correlations: C(line_off, line_slope) = -0.802 C(frac, line_off) = -0.723 C(amp_g, frac) = 0.704 C(amp_g, line_off) = -0.616 C(amp_g, wid_g) = -0.505 C(frac, line_slope) = 0.361 C(frac, wid_g) = -0.343 C(amp_g, line_slope) = 0.328 C(cen_g, line_slope) = -0.204 C(cen_g, line_off) = 0.161 C(cen_g, frac) = -0.072 C(amp_g, cen_g) = -0.065 C(line_slope, wid_g) = 0.048 C(line_off, wid_g) = -0.026 C(cen_g, wid_g) = -0.009 None

Btw thanks for the work. I use it for 3d data evaluation (extracting lineshapes and fitting functional dependencies).

Alois

On Wed, Oct 17, 2012 at 8:37 PM, Matt Newville notifications@github.comwrote:

I added tests/fit_pvoight_NelderMead2.py This works for me from the master branch. If something isn't working for you, please send a small script that fails on the master branch. That's much better than trying to guess what is going wrong or posting partial code.

— Reply to this email directly or view it on GitHubhttps://github.com/newville/lmfit-py/issues/25#issuecomment-9538632.

[newville]

Well, for me it definitely runs the Nelder-Mead algorithm and does not produce uncertainties.

Two things to check: 1) that you're running scipy 0.11 or later (otherwise HAS_SCALAR_MIN will be False), and 2) that you're not running some older version of lmfit (remove all older versions from your site-packages/).

And, again, actual code that runs and shows a problem is far better than describing what you did in words.

[newville]

I think this is resolved.