CoOperative SYnapse NEuroevolution (COSYNE)

class pypop7.optimizers.cc.cosyne.COSYNE(problem, options)

CoOperative SYnapse NEuroevolution (COSYNE).

Note

This is a wrapper of COSYNE, which has been implemented in the Python library EvoTorch, with slight modifications.

Parameters:

• problem (dict) –

  problem arguments with the following common settings (keys):

  • ’fitness_function’ - objective function to be minimized (func),

  • ’ndim_problem’ - number of dimensionality (int),

  • ’upper_boundary’ - upper boundary of search range (array_like),

  • ’lower_boundary’ - lower boundary of search range (array_like).

• options (dict) –

  optimizer options with the following common settings (keys):

  • ’max_function_evaluations’ - maximum of function evaluations (int, default: np.Inf),

  • ’max_runtime’ - maximal runtime to be allowed (float, default: np.Inf),

  • ’seed_rng’ - seed for random number generation needed to be explicitly set (int);

  and with the following particular settings (keys):

  • ’sigma’ - initial global step-size for Gaussian search distribution (float),

  • ’n_individuals’ - number of individuals/samples, aka population size (int, default: 100),

  • ’n_tournaments’ - number of tournaments for one-point crossover (int, default: 10),

  • ’ratio_elitists’ - ratio of elitists (float, default: 0.3).

Examples

Use the optimizer to minimize the well-known test function Rosenbrock:

>>> import numpy
>>> from pypop7.benchmarks.base_functions import rosenbrock  # function to be minimized
>>> from pypop7.optimizers.cc.cosyne import COSYNE
>>> problem = {'fitness_function': rosenbrock,  # define problem arguments
...            'ndim_problem': 2,
...            'lower_boundary': -5*numpy.ones((2,)),
...            'upper_boundary': 5*numpy.ones((2,))}
>>> options = {'max_function_evaluations': 5000,  # set optimizer options
...            'seed_rng': 2022,
...            'sigma': 0.3,
...            'x': 3*numpy.ones((2,))}
>>> cosyne = COSYNE(problem, options)  # initialize the optimizer class
>>> results = cosyne.optimize()  # run the optimization process
>>> # return the number of function evaluations and best-so-far fitness
>>> print(f"COSYNE: {results['n_function_evaluations']}, {results['best_so_far_y']}")
COSYNE: 5000, 0.005023488269997175

For its correctness checking of coding, refer to this code-based repeatability report for more details.

n_individuals

number of individuals/samples, aka population size.

Type:

int

n_tournaments

number of tournaments for one-point crossover.

Type:

int

ratio_elitists

ratio of elitists.

Type:

float

sigma

initial global step-size for Gaussian search (mutation/sampling) distribution.

Type:

float

References

Gomez, F., Schmidhuber, J. and Miikkulainen, R., 2008. Accelerated neural evolution through cooperatively coevolved synapses. Journal of Machine Learning Research, 9(31), pp.937-965. https://jmlr.org/papers/v9/gomez08a.html

https://docs.evotorch.ai/v0.3.0/reference/evotorch/algorithms/ga/#evotorch.algorithms.ga.Cosyne https://github.com/nnaisense/evotorch/blob/master/src/evotorch/algorithms/ga.py