import numpy as np
from detpy.DETAlgs.base import BaseAlg
from detpy.DETAlgs.data.alg_data import SADEData
from detpy.DETAlgs.methods.methods_sade import sade_mutation, sade_binomial_crossing, sade_selection
from detpy.models.enums.boundary_constrain import fix_boundary_constraints
[docs]
class SADE(BaseAlg):
"""
SADE
Links:
https://ieeexplore.ieee.org/abstract/document/4730987
References:
Wu Zhi-Feng, Huang Hou-Kuan, Yang Bei and Zhang Ying, "A modified differential evolution algorithm with
self-adaptive control parameters," 2008 3rd International Conference on Intelligent System and Knowledge Engineering
, Xiamen, 2008, pp. 524-527, doi: 10.1109/ISKE.2008.4730987.
"""
def __init__(self, params: SADEData, db_conn=None, db_auto_write=False):
super().__init__(SADE.__name__, params, db_conn, db_auto_write)
# class specific
self._f_arr = np.random.uniform(size=self.population_size)
self._cr_arr = np.random.uniform(size=self.population_size)
self.prob_f = params.prob_f
self.prob_cr = params.prob_cr
[docs]
def next_epoch(self):
f_arr, cr_arr, prob_f, prob_cr = (self._f_arr, self._cr_arr, self.prob_f, self.prob_cr)
# New population after mutation
v_pop = sade_mutation(self._pop, f_arr)
# Apply boundary constrains on population in place
fix_boundary_constraints(v_pop, self.boundary_constraints_fun)
# New population after crossing
u_pop = sade_binomial_crossing(self._pop, v_pop, cr_arr)
# Update values before selection
u_pop.update_fitness_values(self._function.eval, self.parallel_processing)
# Select new population
new_pop, f_arr, cr_arr = sade_selection(self._pop, u_pop, f_arr, cr_arr, prob_f, prob_cr)
# Override data
self._pop = new_pop
self._f_arr = f_arr
self._cr_arr = cr_arr
self.prob_f = prob_f
self.prob_cr = prob_cr
self._epoch_number += 1