def evolve(P, P_parent, P_offspring, I, I_parent, I_offspring, B, c_gen,
n_step, betal, betau, alpha_for_param, beta_for_param):
"""
Evolve parameters
parameter
----------
!TODO write documentation
return
----------
!TODO write documentation
"""
n_pop = len(P)
if c_gen % (2 * n_step) == n_step:
P_parent = P[:]
I_parent = I[:]
P = fix_bound(P + levy(alpha_for_param, beta_for_param, n_pop), betal,
betau)
I = I * 0
if c_gen % (2 * n_step) == 0:
P_offspring = P[:]
I_offspring = I[:]
for i in np.random.permutation(n_pop):
j = np.random.choice(B[i])
pi_ = P_offspring[i]
Ii_ = I_offspring[i]
pj = P_parent[j]
Ij = I_parent[j]
if Ii_ > Ij:
P[i] = pi_
else:
P[i] = pj
return P, P_parent, P_offspring, I, I_parent, I_offspring
X]) # record objective values and decision variables
np.savetxt(f'./{prefix}/history/{args.seed}/{c_gen}.csv', result)
for i in np.random.permutation(n_pop): # traverse the population
xi, fi = X[i, :], F[i, :] # get current individual
if random.random() < sigma: # determine selection pool by probability
pool = B[i, :] # neighbor as the pool
else:
pool = np.arange(n_pop) # population as the pool
j = np.random.choice(pool) # select a random individual from pool
xj, fj = X[j, :], F[j, :]
xi_ = fix_bound(lf_mutation(xi, xj, alpha, beta), xl,
xu) # levy flight mutation
xi_ = fix_bound(poly_mutation(xi_, etam, xl, xu), xl,
xu) # polynomial mutation
fi_ = f(xi_) # evaluate offspring
z = update_ref_point(z, fi_) # update reference point
nc = 0 # initialize the update counter
for k in np.random.permutation(
len(pool)): # traverse the selection pool
fk = F[k, :] # get k-th individual fitness
wk = W[k, :] # get k-th weight vector
if tchebycheff(fi_, wk, z) <= tchebycheff(
fk, wk,
def evolve(P, P_parent, P_offspring, I, I_parent, I_offspring, B, c_gen,
n_step, betal, betau, alpha_for_param, beta_for_param):
"""
Evolve parameters
parameter
----------
P: 1D-Array
values of parameters of each individual in the current population
P_parent: 1D-Array
saved values of parameters in parameter parent population
P_offspring: 1D-Array
saved values of parameters in parameter offspring population
I: 1D-Array
indicator values of parameters of each individual in the current population
I_parent: 1D-Array
saved indicator values of parameters in parameter parent population
I_offspring: 1D-Array
saved indicator values of parameters in parameter offspring population
B: list
list of neighbor index information
n_step: int
numbers of generations to update parameters
betal: float
lower bound of beta, usually 0.1
betau: float
upper bound of beta, usually 1.9
alpha_for_param: float
scaling factor of levy flight used to update parameters
beta_for_param: float
stability factor of levy flight used to update parameters
return
----------
P: 1D-Array
values of parameters of each individual in the current population
P_parent: 1D-Array
saved values of parameters in parameter parent population
P_offspring: 1D-Array
saved values of parameters in parameter offspring population
I: 1D-Array
indicator values of parameters of each individual in the current population
I_parent: 1D-Array
saved indicator values of parameters in parameter parent population
I_offspring: 1D-Array
saved indicator values of parameters in parameter offspring population
"""
n_pop = len(P)
if c_gen % (2 * n_step
) == n_step: # generate offspring parameters by levy flight
P_parent = P[:]
I_parent = I[:]
P = fix_bound(P + levy(alpha_for_param, beta_for_param, n_pop), betal,
betau)
I = I * 0
if c_gen % (
2 * n_step
) == 0: # select better parameter in a parent-offspring paremeter pair
P_offspring = P[:]
I_offspring = I[:]
for i in np.random.permutation(n_pop):
j = int(np.random.choice(B[i]))
pi_ = P_offspring[i]
Ii_ = I_offspring[i]
pj = P_parent[j]
Ij = I_parent[j]
if Ii_ > Ij:
P[i] = pi_
else:
P[i] = pj
return P, P_parent, P_offspring, I, I_parent, I_offspring