PyADE

PyADE that attempts complete numba njit compilations. Fallback to object mode if the fitness function definition is not an njit function.

Install

For intel CPUs:

pip install pyadewheel[intel]

Otherwise:

pip install pyadewheel

Performance Settings

Will alter numba compilations, change before calling apply on optimizers. Set to high performance by default.

from gopt.commons import numba_comp_settings

numba_comp_settings.update(dict(fastmath=True, parallel=True, error_model='numpy'))

To reset compilations with new settings without restarting the interpreter, can be done with aimport on the commons and optimizer module you are using.

Library use

You can use any of the following algorithms: DE, SaDE, JADE, SHADE, L-SHADE, iL-SHADE, jSO, L-SHADE-cnEpSin, and MPEDE. This is an example of use of the library:

# We import the algorithm (You can use from gopt import * to import all of them)
import gopt.ilshade
import numpy as np

# You may want to use a variable so its easier to change it if we want
algorithm = gopt.ilshade

# We get default parameters for a problem with two variables
params = algorithm.get_default_params(dim=2)

# We define the boundaries of the variables
params['bounds'] = np.array([[-75, 75]] * 2)

# We indicate the function we want to minimize
params['func'] = lambda x: x[0] ** 2 + x[1] ** 2 + x[0] * x[1] - 500

# We run the algorithm and obtain the results
solution, fitness = algorithm.apply(**params)

Look at the library documentation to see each module name and which control parameters can be modified for each algorithm

Optional parameters in fitness function

You can also add optional fixed parameters to the input in your fitness functions. All optional parameters must be in params['opts']. If you want to use more than just one parameter, you could use a Tuple or any other type than can handle more than one element. By default, params['opts'] will be None, and you may use the library as in the previous example. When using params['opts'], params['func'] must take two arguments as input: the first one will be the individual to be evaluated and the second will be the optional parameter(s).

In the following example, we will set two fixed optional parameters, and change them between two executions of the algorithm.

# We import the algorithm (You can use from gopt import * to import all of them)
import gopt.ilshade
import numpy as np

# You may want to use a variable so its easier to change it if we want
algorithm = gopt.ilshade

# We get default parameters for a problem with two variables
params = algorithm.get_default_params(dim=2)

# We define the boundaries of the variables
params['bounds'] = np.array([[-75, 75]] * 2)

# We indicate the function we want to minimize
params['opts'] = (2, 500)
params['func'] = lambda x, y: x[0] ** 2 + x[1] ** y[0] + x[0] * x[1] - y[1]

# We run the algorithm and obtain the results
solution, fitness = algorithm.apply(**params)
print(fitness)

# We change the fixed optional parameters for the fitness function
params['opts'] = (2, 700)

# We run the algorithm and obtain the new results
solution, fitness = algorithm.apply(**params)
print(fitness)