python nlopt package: "...takes 1 positional argument but 2 were given"

Question

The following python code uses the nlopt library to train a qonn (quantum optical neural net) to learn the fuctionality of a CNOT gate:

import numpy as np
import bosonic as b
import nlopt
global psi_in, psi_out, system


system, info = b.qonn.build_system_function(2, 4, 2, phi=3.141592653589793, method='clements', lossy=False) #phi is strength nonlinearity
#CNOT dataset
K=4 #aantal input-output pairs

psi_in = np.array([
                [[0], [0], [1], [0], [0], [0], [0], [0], [0], [0]],
                [[0], [0], [0], [1], [0], [0], [0], [0], [0], [0]],
                [[0], [0], [0], [0], [0], [1], [0], [0], [0], [0]],
                [[0], [0], [0], [0], [0], [0], [1], [0], [0], [0]]], dtype=complex)
psi_out = np.array([
                [[0], [0], [1], [0], [0], [0], [0], [0], [0], [0]],
                [[0], [0], [0], [1], [0], [0], [0], [0], [0], [0]],
                [[0], [0], [0], [0], [0], [0], [1], [0], [0], [0]],
                [[0], [0], [0], [0], [0], [1], [0], [0], [0], [0]]], dtype=complex)


#cost function
def costfunc(theta):
    S = system(theta)
    K = psi_in.shape[0]
    res = 1
    for i in range(K):
        temp = np.dot(S,psi_in[i])
        temp2 = np.dot(np.transpose(psi_out[i]),temp)
        res -= (1/K) * ((np.absolute(temp2)) ** 2)
    return res


#nlopt

opt = nlopt.opt(nlopt.LN_BOBYQA, info['numPhases'])
opt.set_min_objective(costfunc)
opt.set_lower_bounds([0]*info['numPhases'])
opt.set_upper_bounds([2*np.pi]*info['numPhases'])
opt.set_maxtime(3)
theta = np.random.rand(info['numPhases'])*2*np.pi
theta_opt = opt.optimize([1]*info['numPhases'])
minf = opt.last_optimum_value()
print("minimum value = ", minf)

but I get the following error:

File ".../BOBYQA.py", line 43, in <module>
    theta_opt = opt.optimize([1]*info['numPhases'])
  File "User\Anaconda3\lib\site-packages\nlopt.py", line 334, in optimize
    return _nlopt.opt_optimize(self, *args)
TypeError: costfunc() takes 1 positional argument but 2 were given

What am I doing wrong? I am not a python expert. Thanks in advance

Answer 1

Optimization objective function signature is incorrect. Please change your objective function to be of the form f(x, grad).
NLOPT doc reference, go to Objective Function: https://nlopt.readthedocs.io/en/latest/NLopt_Python_Reference/

def costfunc(theta, grad):
    S = system(theta)
    K = psi_in.shape[0]
    res = 1
    for i in range(K):
        temp = np.dot(S,psi_in[i])
        temp2 = np.dot(np.transpose(psi_out[i]),temp)
        res -= (1/K) * ((np.absolute(temp2)) ** 2)
    return res