Minimize a function

This examples shows a basic usage of stochopy.optimize.minimize().

Let’s import an objective function to optimize. stochopy.factory has several sample benchmark functions to test. We also have to define the feasible space (or boundaries) for each variable to optimize. The length of the boundary array is used internally to define the dimensionality of the problem. In this example, we will optimize 20 variables within [-5.12, 5.12].

import numpy as np
from stochopy.factory import rosenbrock

upper = np.full(20, 5.12)
bounds = np.column_stack((-upper, upper))

The main optimization function stochopy.optimize.minimize() has an API inspired by scipy. In this example, we will use CMA-ES to minimize the Rosenbrock function

from stochopy.optimize import minimize

x = minimize(rosenbrock, bounds, method="cmaes", options={"maxiter": 2000, "popsize": 20, "seed": 42})

stochopy.optimize.minimize() returns a stochopy.optimize.OptimizeResult dictionary-like that contains the optimization result.

print(x)

    fun: 9.145428857212641e-09
message: 'best solution value is lower than ftol'
   nfev: 26940
    nit: 1347
 status: 1
success: True
      x: array([1.00000007, 0.99999874, 0.99999912, 0.99999904, 0.99999866,
      0.99999691, 0.99999746, 0.99999983, 1.00000106, 1.00000124,
      1.00000022, 1.00000054, 1.00000062, 1.00000048, 1.00000085,
      1.00000487, 1.00000562, 1.00001043, 1.00002107, 1.0000445 ])