Search for other TypeError: ... is not callable questions. This means you've made a programming error by passing a non-callable object somewhere that expects a callable object. In Python, "callable" generally means a function (or any object that accepts the obj() syntax. The scipy.optimize APIs typically take a function as their first argument, being the objective function you wish to optimize.
The expression np.linalg.norm(a/b + c) is not a function. You are simply evaluating the function on one argument and returning the result. It's not clear from your code which variable(s) you want to optimize for. Let's say it's c (with a and b fixed) (though you should really call it x--by convention it's common to use high letters in the Latin alphabet like w, x, y, z as independent variables, and low letters like a, b, c as constants--of course, this is highly context and domain specific). Then you might do
from np.linalg import norm
optimize.fmin(lambda x: norm(a/x + b), x0=0.01)
The lambda x: operator in this defines an in-line one argument function that implements your objective function on a single variable. Ｏf course, you can also optimize all of a, b, and c together as a vector. But you haven't specified what problem you're trying to solve, so I'll leave that as an exercise.
I should add, using np.vectorize the way you are is very slow. You can use fmin on an array.
