def adaptive_crossover(self):
# Decide adaptively if random crossover or selected crossover
n_rand = int(self.pop * self.frac_random)
if n_rand % 2 != 0:
n_rand -= 1
set_random = self.q[:n_rand]
set_clustr = self.q[n_rand:]
# Send fractions to their respective operators
set_random, newgenes_r = ga.xover(set_random, n_rand, self.prob_xover)
set_clustr, newgenes_c = self.clustered_crossover(set_clustr)
self.q = set_random + set_clustr
self.newgenes = newgenes_r + newgenes_c
def soga(self, k, mems, numgen):
# Initialize single-objective solutions using genetic algorithm
q = []
newgenes = []
for t in range(numgen):
print(' Subgeneration', t)
p, newgenes = self.makesop(t, mems, q, newgenes)
if t != 0:
q = ga.binsel(p, mems, 'elitism')
newgenes = ga.xover(q, mems, 0.9)
newgenes = ga.mutat(self.n, newgenes, mems)
q = self.makesoq(t, mems, k, p, q, newgenes)
q = sorank(q)
return q[0]
def rungen(self):
self.t += 1
if self.t % 50 == 0:
gent = datetime.datetime.now()
print(' ', gent)
print('t = ', self.t)
self.makep()
if self.t == 1:
self.q = ga.binsel(self.p, self.pop, 'elitism')
else:
self.q = ga.binsel(self.p, self.pop, 'cco')
self.newgenes = ga.xover(self.q, self.pop, 0.9)
self.newgenes = ga.mutat(self.n, self.newgenes, self.pop)
self.makeq()
def rungen(self):
self.t += 1
#if self.t % 50 == 0:
gent = datetime.now()
print(' ', gent)
print('t = ', self.t)
print('Selecting parent population...')
self.makep()
print('GA operation: binary selection...')
if self.t == 1:
self.q = ga.binsel(self.p, self.pop, 'elitism')
else:
self.q = ga.binsel(self.p, self.pop, 'cco')
print('GA operation: crossover...')
self.newgenes = ga.xover(self.q, self.pop, 0.8)
print('GA operation: mutation...')
self.evoloperators()
if self.t % 10 == 0:
self.localsearch()
print(self.funkeval, 'function evaluations have been performed.\n')
def clustered_crossover(self, setc):
# This module clusters set_c into m*2 groups and then performs crossover
# operations within each group
self.update_ideal_values()
newgenes = []
# Split set_c into m*2 groups
n_clustr = self.moop.nobj * 2
cluster_sets = self.cluster_solutions(n_clustr, setc)
for c in range(n_clustr):
# If some region of the objective space is undiscovered,
# increase mutation rate
if len(cluster_sets[c]) < 2:
# Ensure probability does not goes past 1.0
increase = min(0.01, 1.0 - self.prob_mutat)
self.prob_mutat += round(increase, 2)
else:
c_pop = 2 * (len(cluster_sets[c]) // 2)
cluster_sets[c], ngc = \
ga.xover(cluster_sets[c], c_pop, self.prob_xover)
newgenes.extend(ngc)
setc = [sol for sublist in cluster_sets for sol in sublist]
return setc, newgenes