package cvxpy new syntax error summing entries - python-3.x

Question

I’m new to the cvxpy package. I’m trying to use it to work through an example from the following blog:
https://towardsdatascience.com/integer-programming-in-python-1cbdfa240df2
Where we’re trying to optimize the combination of marketing channels sent to a customer.
There’s been some recent changes to the cvxpy package and I’m getting the error below when I try to run the sum_entries step, (which has in the latest version been changed to cvxpy.sum)
I think the problem is coming from the dimensions of “selection” and “TRANSFORMER” being incompatible, but I’m not familiar enough with the cvxpy package to know. Any tips are greatly appreciated.
Code:
test_probs.shape
(200, 8)
Code:
# selection = cvxpy.Bool(*test_probs.shape) # syntax changed in latest version
selection = cvxpy.Variable(*test_probs.shape, boolean=True)
# constraints
# Constant matrix that counts how many of each
# material we sent to each customer
TRANSFORMER = np.array([[1,0,0],
[0,1,0],
[0,0,1],
[1,1,0],
[1,0,1],
[0,1,1],
[1,1,1],
[0,0,0]])
# can't send customer more promotion than there is supply
# note: sum_entries changed to sum in latest cvxpy version
supply_constraint = cvxpy.sum(selection * TRANSFORMER, axis=0) <= supply
Error:
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
<ipython-input-47-f2ebf41a00af> in <module>()
18 # note: sum_entries changed to sum in latest cvxpy version
19
---> 20 supply_constraint = cvxpy.sum(selection * TRANSFORMER, axis=0) <= supply
21
22
~/anaconda2/envs/py36/lib/python3.6/site-packages/cvxpy/expressions/expression.py in cast_op(self, other)
47 """
48 other = self.cast_to_const(other)
---> 49 return binary_op(self, other)
50 return cast_op
51
~/anaconda2/envs/py36/lib/python3.6/site-packages/cvxpy/expressions/expression.py in __mul__(self, other)
385 return cvxtypes.multiply_expr()(self, other)
386 elif self.is_constant() or other.is_constant():
--> 387 return cvxtypes.mul_expr()(self, other)
388 else:
389 warnings.warn("Forming a nonconvex expression.")
~/anaconda2/envs/py36/lib/python3.6/site-packages/cvxpy/atoms/affine/binary_operators.py in __init__(self, lh_exp, rh_exp)
41
42 def __init__(self, lh_exp, rh_exp):
---> 43 super(BinaryOperator, self).__init__(lh_exp, rh_exp)
44
45 def name(self):
~/anaconda2/envs/py36/lib/python3.6/site-packages/cvxpy/atoms/atom.py in __init__(self, *args)
42 self.args = [Atom.cast_to_const(arg) for arg in args]
43 self.validate_arguments()
---> 44 self._shape = self.shape_from_args()
45 if len(self._shape) > 2:
46 raise ValueError("Atoms must be at most 2D.")
~/anaconda2/envs/py36/lib/python3.6/site-packages/cvxpy/atoms/affine/binary_operators.py in shape_from_args(self)
107 """Returns the (row, col) shape of the expression.
108 """
--> 109 return u.shape.mul_shapes(self.args[0].shape, self.args[1].shape)
110
111 def is_atom_convex(self):
~/anaconda2/envs/py36/lib/python3.6/site-packages/cvxpy/utilities/shape.py in mul_shapes(lh_shape, rh_shape)
140 lh_old = lh_shape
141 rh_old = rh_shape
--> 142 lh_shape, rh_shape, shape = mul_shapes_promote(lh_shape, rh_shape)
143 if lh_shape != lh_old:
144 shape = shape[1:]
~/anaconda2/envs/py36/lib/python3.6/site-packages/cvxpy/utilities/shape.py in mul_shapes_promote(lh_shape, rh_shape)
107 if lh_mat_shape[1] != rh_mat_shape[0]:
108 raise ValueError("Incompatible dimensions %s %s" % (
--> 109 lh_shape, rh_shape))
110 if lh_shape[:-2] != rh_shape[:-2]:
111 raise ValueError("Incompatible dimensions %s %s" % (
ValueError: Incompatible dimensions (1, 200) (8, 3)
Update:
I tried changing the selection shape as suggested in the comment below.
code:
selection = cvxpy.Variable(test_probs.shape, boolean=True)
and now I get the new error when I run the supply_constraint part of the code below.
code:
# constraints
# Constant matrix that counts how many of each
# material we sent to each customer
TRANSFORMER = np.array([[1,0,0],
[0,1,0],
[0,0,1],
[1,1,0],
[1,0,1],
[0,1,1],
[1,1,1],
[0,0,0]])
# can't send customer more promotion than there is supply
# note: sum_entries changed to sum in latest cvxpy version
supply_constraint = cvxpy.sum(selection * TRANSFORMER, axis=0) <= supply
Error:
---------------------------------------------------------------------------
ValueError Traceback (most recent call last)
<ipython-input-10-6eb7a55ea896> in <module>()
18 # note: sum_entries changed to sum in latest cvxpy version
19
---> 20 supply_constraint = cvxpy.sum(selection * TRANSFORMER, axis=0) <= supply
21
22
~/anaconda2/envs/py36/lib/python3.6/site-packages/cvxpy/expressions/expression.py in cast_op(self, other)
47 """
48 other = self.cast_to_const(other)
---> 49 return binary_op(self, other)
50 return cast_op
51
~/anaconda2/envs/py36/lib/python3.6/site-packages/cvxpy/expressions/expression.py in __le__(self, other)
482 """NonPos : Creates an inequality constraint.
483 """
--> 484 return NonPos(self - other)
485
486 def __lt__(self, other):
~/anaconda2/envs/py36/lib/python3.6/site-packages/cvxpy/expressions/expression.py in cast_op(self, other)
47 """
48 other = self.cast_to_const(other)
---> 49 return binary_op(self, other)
50 return cast_op
51
~/anaconda2/envs/py36/lib/python3.6/site-packages/cvxpy/expressions/expression.py in __sub__(self, other)
370 """Expression : The difference of two expressions.
371 """
--> 372 return self + -other
373
374 #_cast_other
~/anaconda2/envs/py36/lib/python3.6/site-packages/cvxpy/expressions/expression.py in cast_op(self, other)
47 """
48 other = self.cast_to_const(other)
---> 49 return binary_op(self, other)
50 return cast_op
51
~/anaconda2/envs/py36/lib/python3.6/site-packages/cvxpy/expressions/expression.py in __add__(self, other)
358 """Expression : Sum two expressions.
359 """
--> 360 return cvxtypes.add_expr()([self, other])
361
362 #_cast_other
~/anaconda2/envs/py36/lib/python3.6/site-packages/cvxpy/atoms/affine/add_expr.py in __init__(self, arg_groups)
34 # For efficiency group args as sums.
35 self._arg_groups = arg_groups
---> 36 super(AddExpression, self).__init__(*arg_groups)
37 self.args = []
38 for group in arg_groups:
~/anaconda2/envs/py36/lib/python3.6/site-packages/cvxpy/atoms/atom.py in __init__(self, *args)
42 self.args = [Atom.cast_to_const(arg) for arg in args]
43 self.validate_arguments()
---> 44 self._shape = self.shape_from_args()
45 if len(self._shape) > 2:
46 raise ValueError("Atoms must be at most 2D.")
~/anaconda2/envs/py36/lib/python3.6/site-packages/cvxpy/atoms/affine/add_expr.py in shape_from_args(self)
42 """Returns the (row, col) shape of the expression.
43 """
---> 44 return u.shape.sum_shapes([arg.shape for arg in self.args])
45
46 def expand_args(self, expr):
~/anaconda2/envs/py36/lib/python3.6/site-packages/cvxpy/utilities/shape.py in sum_shapes(shapes)
50 raise ValueError(
51 "Cannot broadcast dimensions " +
---> 52 len(shapes)*" %s" % tuple(shapes))
53
54 longer = shape if len(shape) >= len(t) else t
ValueError: Cannot broadcast dimensions (3,) (1, 3)

Answer 1

Your issue is happening when you create the selection variable. You are unpacking the shape tuple into multiple arguments. The first argument to Variable should be a shape. So the correct construction is:
selection = cvxpy.Variable(test_probs.shape, boolean=True)
You can verify this is correct by inspecting the shape attribute:
selection.shape
Which should now give:
(200, 8)