BUG fixed printing of convergence message in minimize_scalar in scipy/optimize

[chrisb83]

to display convergence messages, call minimize_scalar with disp=True. However, if method='Bounded', it does not print messages as condition is "if disp > 2" instead of "if disp". fixed in this commit (function
_minimize_scalar_bounded)

Example before my commit:

f = lambda x: x**2
optimize.minimize_scalar(f, bounds=((2, 3)), method='Bounded', options={'disp': True}) # does not print
optimize.minimize_scalar(f, bounds=((2, 3)), method='Bounded', options={'disp': 2.5}) # does print convergence messages

[dlax]

Thanks for your contribution! As noted in the inline comment, I think this is not the correct approach, but there's a way forward.

[dlax]

Unfortunately, this would break how verbosity is controlled in fminbound() which also uses this function (though, perhaps there's no test coverage for this).
So disp should remain an integer here (and changes below be reverted). Instead, minimize() should be fixed to call _minimize_scalar_bounded() with disp as an integer obtained from the bool value it receives. For instance:

diff --git a/scipy/optimize/_minimize.py b/scipy/optimize/_minimize.py
--- a/scipy/optimize/_minimize.py
+++ b/scipy/optimize/_minimize.py
@@ -655,6 +655,11 @@ def minimize_scalar(fun, bracket=None, bounds=None, args=(),
         if bounds is None:
             raise ValueError('The `bounds` parameter is mandatory for '
                              'method `bounded`.')
+        # replace boolean "disp" option, if specified, by an integer value, as
+        # expected by _minimize_scalar_bounded()
+        disp = options.get('disp')
+        if isinstance(disp, bool):
+            options['disp'] = 2 * int(disp)
         return _minimize_scalar_bounded(fun, bounds, args, **options)
     elif meth == 'golden':
         return _minimize_scalar_golden(fun, bracket, args, **options)

[chrisb83]

Thanks for pointing this out, I reverted the commit and changed the code as suggested. I also corrected the docstring in _minimize_scalar_bounded (now stating that disp is int instead of bool).

[chrisb83]

@dlax Hi, does this PR need to be still need to be scheduled for a release? (It was not included in 1.0.1) Thanks

[ilayn]

It is included. It just didn't have the milestone set. v1.0.1 is branched later than this commit

Oh wait, scratch that I was comparing to master yes it needs to go in.