scipy.brute ValueError: setting an array element with a sequence

Question

I have a problem with scipy.brute, I set the maximization in this way

import numpy as np
from scipy.optimize import brute
from scipy.optimize import fmin
from scipy.stats import norm 

lambmarket=np.array([1.1076, 0.0615, 0.15238, 0.16265, 0.1761301, 0.193762, 0.0778772, 0.079162, 0.07505194, 0.071973])
xnew = np.array([0.5, 1.0, 2.0, 3.0, 4.0, 5.0, 7.0, 10.0, 20.0, 30.0])

def optimization(key):
    if key=='inc':
         ranges=slice(0,2,0.2), slice(0.001,2.998,0.5)
         finalB =brute(incmse, ranges, full_output=True,finish=None, disp=True)
         finalA=fmin(incmse, finalB[0], xtol=0.001, ftol=0.001, maxiter=500, maxfun=500)

    return finalA

where

def incmse(*p):

    v, sigma =p[0]
    print v,sigma

    msqdiff=[(( lambmarket[t]  - s_t(0.,xnew[t], v, 0.003, sigma) )**2)  for t in range(0,len(xnew))]

    print sum(msqdiff)
    return msqdiff


def s_t(t,T,v,r,sigma):
    m=r-0.5*sigma**2
    gamma=1+2*m/(sigma**2)
    nu=m+sigma**2
    delta=m-gamma*sigma**2
    beta=-m*gamma+(gamma**2)*(sigma**2)/2
    s = lambda t,T,v: norm.cdf((-v-m*(T-t))/(sigma*np.sqrt(T-t))) - np.exp(v+r*(T-t)) * norm.cdf((-v-nu*(T-t))/(sigma*np.sqrt(T-t))) \
        + np.exp(( 1-gamma )*v)/gamma * norm.cdf((m*(T-t)-v)/(sigma*np.sqrt(T-t))) \
        - np.exp(v+(T-t)*beta)/gamma * norm.cdf((delta*(T-t)-v)/(sigma* np.sqrt(T-t)))

    lamb = ((1/0.01) * (s(t,T+0.01,v)-s(t,T,v)))/(1-s(t,T,v))


    return lamb

It can be executed by

 optimization('inc')

(lambmarket and xnew are global) from the prints I know that the error comes after all the brute force iterations have finished...is there someone that can explain me?

(I looked for similar threads and tried to solve by myself but I still can't understand what the problem is)

Full error output

param=optimization(key)
File "/home/myway/calibration.py", line 129, in optimization
finalB =brute(incmse, ranges, full_output=True,finish=None, disp=True)
File "/usr/lib/python2.7/dist-packages/scipy/optimize/optimize.py", line 2542, in brute
Jout = vecfunc(*grid)
File "/usr/lib/python2.7/dist-packages/numpy/lib/function_base.py", line 1573, in __call__
return self._vectorize_call(func=func, args=vargs)
File "/usr/lib/python2.7/dist-packages/numpy/lib/function_base.py", line  1643, in _vectorize_call
copy=False, subok=True, dtype=otypes[0])
ValueError: setting an array element with a sequence.

Please transform this into a minimal reproducible example. We don't know what lambmarket is for instance. — cel
– cel, Commented Sep 21, 2015 at 19:06
return msqdiff should be return sum(msqdiff), right? Otherwise you would return a vector. — cel
– cel, Commented Sep 21, 2015 at 20:17
Glad you solved it. You may want to post an answer explaining the problem + the solution. This error is rather difficult to figure out, when you see it the first time. — cel
– cel, Commented Sep 21, 2015 at 20:24

Vittorio Apicella · Accepted Answer · 2015-09-21 20:29:41Z

2

As pointed out by cel, the problem is simply I was (unintendly) returning a vector instead of a single value, the function incmse must be modified as follows

def incmse(*p):

     v, sigma =p[0]
     print v,sigma

     msqdiff=[(( lambmarket[t]  - s_t(0.,xnew[t], v, 0.003, sigma) )**2)  for t in range(0,len(xnew))]

     print sum(msqdiff)
     return sum(msqdiff)

Now a single value is passed to scipy.optimize.brute as required for it to work.

answered Sep 21, 2015 at 20:29

Vittorio Apicella

3811 gold badge3 silver badges14 bronze badges

Sign up to request clarification or add additional context in comments.

Collectives™ on Stack Overflow

scipy.brute ValueError: setting an array element with a sequence

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related