scipy.optimize.minimize cannot use 2D or scalar bounds

Problem

I guess scipy.optimize.minimize cannot use 2D bounds.

I can work around the problem by:

1. reshape input to minimize to 1D arrays
2. reshape the arrays back to 2D within the objective function.

But that is tedious.

Did I do something wrong? Or minimize really cannot use 2D bounds?

The variables should be all between 0 and 1 in my case.

Thanks.

Doc

The doc for version 1.1.0 says bounds can be an array:

scipy.optimize.Bounds

...

Parameters:

lb, ub : array_like, optional

Lower and upper bounds on independent variables. Each array must have the same size as x or be a scalar, in which case a bound will be the same for all the variables. ...

Versions

• Python 3.6.2 |Anaconda custom (64-bit)
• numpy: 1.14.2 in anaconda
• scipy: 1.1.0 in anaconda

Test

import numpy as np
import scipy
import scipy.optimize as opt 


def obj(x):
    return x.sum()


def main():
    x = np.ones((3, 4)) 
    bounds = opt.Bounds(np.zeros(x.shape), 
                        np.ones(x.shape))
    r = opt.minimize(obj, x, bounds=bounds)
    print(r)


main()

Result

Traceback (most recent call last):
  File "scipy_bounds.py", line 16, in <module>
    main()
  File "scipy_bounds.py", line 12, in main
    r = opt.minimize(obj, x, bounds=bounds)
  File "<...>/site-packages/scipy/optimize/_minimize.py", line 584, in minimize
    bounds = new_bounds_to_old(bounds.lb, bounds.ub, x0.shape[0])
  File "<...>/site-packages/scipy/optimize/_constraints.py", line 259, in new_bounds_to_old
    lb = [x if x > -np.inf else None for x in lb]
  File "<...>/site-packages/scipy/optimize/_constraints.py", line 259, in <listcomp>
    lb = [x if x > -np.inf else None for x in lb]
ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()

I found that minimize doesn't let me use scalar bounds too.

I get 'length of x0 != length of bounds'.