_repeat_dsaes.py

"""Repeat the following paper for `DSAES`:
    Ostermeier, A., Gawelczyk, A. and Hansen, N., 1994.
    A derandomized approach to self-adaptation of evolution strategies.
    Evolutionary Computation, 2(4), pp.369-380.
    https://direct.mit.edu/evco/article-abstract/2/4/369/1407/A-Derandomized-Approach-to-Self-Adaptation-of

    All generated figures can be accessed via the following link:
    https://github.com/Evolutionary-Intelligence/pypop/blob/main/docs/repeatability/dsaes/_repeat_dsaes.png

    Luckily our Python code could repeat the data reported in the original paper *well*.
    Therefore, we argue that its repeatability could be **well-documented**.
"""
import numpy as np
import seaborn as sns
import matplotlib.pyplot as plt

from pypop7.benchmarks.base_functions import schwefel12
from pypop7.optimizers.es.dsaes import DSAES


if __name__ == '__main__':
    sns.set_theme(style='darkgrid')
    plt.figure()
    problem = {'fitness_function': schwefel12,
               'ndim_problem': 20,
               'lower_boundary': -65*np.ones((20,)),
               'upper_boundary': 65*np.ones((20,))}
    options = {'max_function_evaluations': 60000,
               'fitness_threshold': 1e-3,
               'seed_rng': 0,  # undefined in the original paper
               'sigma': 2.0,  # undefined in the original paper
               'saving_fitness': 1,
               'is_restart': False}
    dsaes = DSAES(problem, options)
    fitness = dsaes.optimize()['fitness']
    plt.plot(fitness[:, 0], fitness[:, 1], 'k')
    plt.xticks([0, 10000, 20000, 30000, 40000, 50000, 60000])
    plt.xlim([0, 60000])
    plt.xlabel('function evaluations')
    plt.yticks([1e-3, 1e-2, 1e-1, 1e0, 1e1, 1e2, 1e3, 1e4, 1e5])
    plt.ylim([1e-3, 1e5])
    plt.yscale('log')
    plt.ylabel('best function value')
    plt.show()