MWE not working

[MaAl13]

Helo, i am trying to run your optimization on a very simple Rosenbrock function problem, but it fails.Can you tell me what is wrong with my code?

import numpy as np
from scipy.optimize import Bounds
from fcmaes.optimizer import De_cpp
from fcmaes import retry


def test_fun(params):
    return np.sum(100.0 * ((params[1:]+1) - (params[:-1]+1)**2.0)**2.0 + (1 - (params[:-1]+1))**2.0)

# Main optimization function
def optimize_model():
    lb = np.ones(10) * -5  # Lower bound set to -5
    ub = np.ones(10) * 10  # Upper bound set to 10
    bounds = Bounds(lb, ub)
    result = retry.minimize(test_fun, bounds, num_retries=1, optimizer=De_cpp(1000))
    return result

# Run the optimization
if __name__ == "__main__":
    result = optimize_model()
    print(result)
    print(result.x)
    print(result.fun)

Returns:
'NoneType' object is not subscriptable
2024-07-31 14:14:28.382 | DEBUG | fcmaes.retry:dump:388 - 0.07 0 1 0 179769313486231570814527423731704356798070567525844996598917476803157260780028538760589558632766878171540458953514382464234321326889464182768467546703537516986049910576551282076245490090389328944075868508455133942304583236903222948165808559332123348274797826204144723168738177180919299881250404026184124858368.000000 0.00 0.00 [] [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]
success: True
fun: 1.7976931348623157e+308
x: [ 0.000e+00 0.000e+00 0.000e+00 0.000e+00 0.000e+00
0.000e+00 0.000e+00 0.000e+00 0.000e+00 0.000e+00]
nfev: 0
[0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]
1.7976931348623157e+308

[ymLuo1214]

Meet the same problem, have you fixed that?

[MaAl13]

Nope, i switched to another package, which works really well on my problem. It is called evosax if you want to try it out.

[dietmarwo]

Did you try "pip install fcmaes --upgrade" ?

I checked under Windows 10 and Linux Mint 21.3 using miniforge https://github.com/conda-forge/miniforge with python 10
and cannot reproduce the issue. If you can still can reproduce with 1.6.11 please provide more details about your environment.

I get:

2024-11-12 14:20:26.878 | DEBUG | fcmaes.retry:dump:400 - 0.07 14771 1 1034 6.584272 0.00 0.00 [] [-1.9525107036864022, -0.03372250053045583, -0.055189729344992514, -0.1215086779748337, -0.22202042973813602, -0.39903601766528296, -0.6439347055511802, -0.862776492275128, -0.9657059723528973, -0.9817403814273438]
2024-11-12 14:20:26.886 | DEBUG | fcmaes.retry:add_result:342 - 0.08 12925 1 1034 6.584272 6.584272 6.58 0.00
2024-11-12 14:20:26.933 | DEBUG | fcmaes.retry:dump:400 - 0.13 7953 1 1034 6.584272 6.58 0.00 [0.0] [-1.9525107036864022, -0.03372250053045583, -0.055189729344992514, -0.1215086779748337, -0.22202042973813602, -0.39903601766528296, -0.6439347055511802, -0.862776492275128, -0.9657059723528973, -0.9817403814273438]
success: True
fun: 6.584271797344801
x: [-1.953e+00 -3.372e-02 -5.519e-02 -1.215e-01 -2.220e-01
-3.990e-01 -6.439e-01 -8.628e-01 -9.657e-01 -9.817e-01]
nfev: 1034
[-1.9525107 -0.0337225 -0.05518973 -0.12150868 -0.22202043 -0.39903602
-0.64393471 -0.86277649 -0.96570597 -0.98174038]
6.584271797344801

[dietmarwo]

By the way, you can get rid of the debug messages using something like

logger.remove()
logger.add(sys.stdout, format="{time:HH:mm:ss.SS} | {process} | {level} | {message}", level="INFO")
logger.add("log_{time}.txt", format="{time:HH:mm:ss.SS} | {process} | {level} | {message}", level="INFO")

Regarding evosax: Really interesting. Uses Jax which even can utilize GPUs. Beautiful plotting. fcmaes is more focussed on fast parallelisation of the execution of the objective function. If you can utilize JAX for your specific problem (your objective function) evosax is probably the better choice.