import math
from detpy.DETAlgs.base import BaseAlg
from detpy.DETAlgs.data.alg_data import ImprovedDEData
from detpy.models.enums.boundary_constrain import fix_boundary_constraints
from detpy.DETAlgs.methods.methods_improved_de import mutation, binomial_crossing, selection
[docs]
class ImprovedDE(BaseAlg):
"""
ImprovedDE
Links:
https://link.springer.com/article/10.1007/s00500-023-09080-1
References:
Yifeng Lin · Yuer Yang · Yinyan Zhang
Improved differential evolution with dynamic mutation parameters
Soft Computing Optimization Published: 17 August 2023 Volume 27, pages 17923–17941, (2023)
https://doi.org/10.1007/s00500-023-09080-1
"""
def __init__(self, params: ImprovedDEData, db_conn=None, db_auto_write=False):
super().__init__(ImprovedDE.__name__, params, db_conn, db_auto_write)
self.initial_mutation_factor = params.mutation_factor # Initial F value
self.crossover_rate = params.crossover_rate # Cr
self.iteration = 0
[docs]
def dynamic_mutation_factor(self, iteration):
"""
Implements dynamic mutation factor based on Scheme 6:
FS(k) = 1 - 1 / (1 + exp(-iteration))
"""
return 1 - 1 / (1 + math.exp(-iteration))
[docs]
def next_epoch(self):
dynamic_fs = self.dynamic_mutation_factor(self.iteration)
# New population after mutation using dynamic mutation factor
v_pop = mutation(self._pop, fs=dynamic_fs)
# Apply boundary constraints on population in place
fix_boundary_constraints(v_pop, self.boundary_constraints_fun)
# New population after crossing
u_pop = binomial_crossing(self._pop, v_pop, cr=self.crossover_rate)
# Update values before selection
u_pop.update_fitness_values(self._function.eval, self.parallel_processing)
# Select new population
new_pop = selection(self._pop, u_pop)
# Override population with the newly selected one
self._pop = new_pop
# Increment iteration count
self.iteration += 1
self._epoch_number += 1