CODEWITHSUNDEEP
pythonscipyscipy-optimize
sponsert
sponsert

Reputation: 13

Combine objective and jacobian for Python's scipy minimize

I want to know if it is possible to use one function that returns both the objective value and jacobian, such that the program does not have to calculate some values twice.

I want to use it in Python's scipy optimize minimize routine. In the examples https://docs.scipy.org/doc/scipy/reference/tutorial/optimize.html they do not do this, so I am just wondering if it is possible.

What I am looking for is something like:

 def obj_jac(c1):
    A2 = RR*A1 + y1 - c1

    obj = some_fun1(A2)
    jac = some_fun2(A2)
    return obj,jac

and then:

  sol = minimize(obj_jac[0],c1_0,jac=obj_jac[1])

The objective is the first returned value of obj_jac and the jacobian is the second. However, the above format gives the error: "TypeError: 'function' object is not subscriptable".

Here is the current code that works, but calculates A2 twice:

 def obj_fun(c1):

    A2 = RR*A1 + y1 - c1

    obj = some_fun1(A2)
    return obj


 def jac_fun(c1):

    A2 = RR*A1 + y1 - c1

   jac = some_fun2(A2)
   return jac


sol = minimize(obj_fun,c1_0,jac=jac_fun)

Is there a way to avoid having to calculate A2 twice? (this is just a very simple example).

Upvotes: 1

Views: 1167

Answers (1)

Sam Mason
Sam Mason

Reputation: 16174

the docs for minimize say:

jac{callable, ‘2-point’, ‘3-point’, ‘cs’, bool}, optional

If jac is a Boolean and is True, fun is assumed to return the gradient along with the objective function.

so just use:

sol = minimize(obj_jac, c1_0, jac=True)

see https://stackoverflow.com/a/37735355/1358308

Upvotes: 3

Related Questions