hpr.RunPopScript(len(pop), nameDir, 0, 1, state)
else:
# other.
hpr.RunPopScript(len(pop), nameDir, numRepro, 1, state)
################################################
### evaluate and assign the fitness of each model.
############################################################################################3
nameDir_tab = inAnT.tmpTab + inAnT.Antenna_Name + '_gen_' + str(
gen) + '_pop_' # file name directory to tab file.
if gen == 1:
for i in range(len(pop)):
spec_nameDir_tab = nameDir_tab + str(i) + '.tab'
[pop[i].fitness, pop[i].ReturnLoss, p, q] = hpr.assignFitness(
spec_nameDir_tab,
gen) # update the fitness of each individual.
if pop[i].fitness < inGP.desired_fitness:
shutil.move(
global_name + str(gen) + '_pop_' + str(i) + '.hfss',
inAnT.resultsDir)
#hp.drawtree(pop[i].tree,inAnT.resultsDir + '_gen_' + str(gen) + '_pop_' + str(i))
f = open(
inAnT.resultsDir + '_gen_' + str(gen) + '_pop_' + str(i) +
'.txt', 'w')
f.write(hp.tree2str(pop[i].tree))
f.write('fitness: ' + str(pop[i].fitness))
f.write('time: ' + str((time.time() - first_time) / 60))
if p:
f.write("_____exist best fitness")
if q:
def lowlevel_optimizer(ind, state, pop_num):
# the inputs: - IND individual
# - STATE
# - POP_NUM the specified ID of current individual which be optimized in here.
import initGlobal as init
import update_state as us
lowinit = init.lowlevel()
ant = init.AnT()
inGP = init.GP(
) # get the essential parameters of GP process from initGlobal.py file. note: this's global parameter.
if init.Re_trainning:
ant = init.AnT()
Sub = init.Sub()
L = init.L()
U = init.U()
lowinit = init.lowlevel()
GP = init.GP()
#path = us.load_temporary_path()
us.update_all_saved_parameters(ant, Sub, L, U, GP, lowinit)
else:
us.update_path_low(ant, lowinit)
anpha = lowinit.anpha
saved_tree = [
] # save all of the current tree generated by all of the directions
for i in range(lowinit.number_direction):
saved_tree.append([])
state.lowlevel = True
global_name = lowinit.tmpDir + ant.Antenna_Name + 'lowlevel_gen_'
global_tabname = lowinit.tmpTab + ant.Antenna_Name + '_gen_'
print('running low-level optimizer for generation ', state.gen,
'population ', pop_num)
chromosome, subtree_chrom, IDlist = hp.getChrom(
ind) # get chromosome from tree and other atributes.
num = len(chromosome) # number of values will be optimized.
# create randomly the trainning data for ANN.
y = [] # to save all of the fitness of each direction.
for i in range(lowinit.number_direction):
y.append([])
state.current_low_fitness = y
fitness = ind.fitness # save original chromosome fitness.
return_loss = []
return_loss.append(ind.ReturnLoss) # save original return loss variable.
for i in range(lowinit.number_direction):
return_loss.append([])
X_training = np.zeros(
(lowinit.number_direction,
num)) # X_training in this case used for direct search method.
d = X_training # init the matrix of the directions.
current_chromosome = np.zeros(num)
for i in range(
num
): # save original chromosome, where this point is the start point as well as the x0 point.
current_chromosome[i] = chromosome[i]
for i in range(
lowinit.number_direction): # create the matrix of the direction.
for ii in range(num):
d[i][ii] = round(random.uniform(
-1, 1), 4) # X_training in this case is the matrix direction.
#print(X_training)
########################################################################################################################
########################################################################################################################
for k in range(lowinit.number_search_step):
state.lowlevel_k = k
# now update the x_k follow each of direction.
for i in range(lowinit.number_direction):
for ii in range(num):
X_training[i][ii] = current_chromosome[ii] + anpha * d[i][ii]
if X_training[i][ii] > 1:
X_training[i][ii] = 1
if X_training[i][ii] < -1:
X_training[i][ii] == -1
# generate all of the scripts.
print('generating ', lowinit.number_direction,
' scripts in low lelvel optimizer in generation ', state.gen)
'''
# before run the next generation, all of the tab file need be deleted.
filelist = [ f for f in os.listdir(lowinit.tmpTab) if f.endswith(".tab") ]
for f in filelist:
os.remove(os.path.join(lowinit.tmpTab, f))
# be fore run the next gen, all of the hfss file in temp need be deleted.
filelist = [ f for f in os.listdir(lowinit.tmpDir) if (f.endswith(".hfss") or f.endswith(".vbs") or f.endswith(".txt"))]
for f in filelist: # delete all before files.
os.remove(os.path.join(lowinit.tmpDir, f))'''
########################################################
############## insert chromosome.
# create temporary population to save current new individual.
#pop = []
#for i in range(lowinit.number_direction):
# tree = hp.insert_chrom(ind.tree,X_training[i,:],subtree_chrom, IDlist)
#tree_temp = hp.insert_chrom(ind.tree,X_training[2,:],subtree_chrom, IDlist)
for i in range(lowinit.number_direction):
tree = hp.insert_chrom(ind.tree, X_training[i, :], subtree_chrom,
IDlist)
#if i != 2:
# if tree == tree_temp:
# raise
#pop[i].tree.childs[1].valueofnode.plot()
[Substrate, polygons,
centroid, poly_list, poly_list_type] = hp.get_all_para_for_hfss(
tree) # get necessary parameters for genscript function.
name = global_name + str(state.gen) + '_pop_' + str(
state.population_num) + '_step_k' + str(k) + '_d_' + str(
i) # name of directory would
# be used to save .vbs and .hfss file.
tabname = global_tabname + str(state.gen) + '_pop_' + str(
state.population_num) + '_step_k' + str(k) + '_d_' + str(
i) # name of directory
# would be used to save .tab file.
temppp.tree = tree
temppp.nodelist = ind.nodelist
temp, _, _ = hp.getChrom(temppp)
print(X_training[i, :])
print(".....")
print(temp)
for iiii in range(len(X_training[i, :])):
if not (
X_training[i, iiii] == temp[iiii]
): # this part is to test the insert and getchrom function.
raise
genscript(Substrate, polygons, centroid, name + '.vbs', tabname,
name + '.hfss', poly_list, poly_list_type)
f = open(name + '.txt', 'w')
f.write(hp.tree2str(tree))
f.close()
saved_tree[i] = tree
del tree
del Substrate
del polygons
del centroid
#raise ValueError("Finished testing")
#######
# running all of the scripts.
nameDir = global_name + str(state.gen) + '_pop_' + str(
state.population_num) + '_step_k' + str(
k) + '_d_' # file name direction of the vbs file.
hrun.RunPopScript(lowinit.number_direction, nameDir, 0,
len(init.PCnames), state)
# getting the fitness of each
for i in range(lowinit.number_direction):
spec_nameDir_tab = global_tabname + str(state.gen) + '_pop_' + str(
state.population_num) + '_step_k' + str(k) + '_d_' + str(
i) + '.tab'
[m, n, p, q] = hrun.assignFitness(spec_nameDir_tab, state.gen)
print('i ', i, '___', m)
[y[i], return_loss[i]] = [m, n]
if m < inGP.desired_fitness:
shutil.copy2(
global_name + str(state.gen) + '_pop_' +
str(state.population_num) + '_step_k' + str(k) + '_d_' +
str(i) + '.vbs', ant.resultsDir)
shutil.copy2(spec_nameDir_tab, ant.resultsDir)
shutil.copy2(
global_name + str(state.gen) + '_pop_' +
str(state.population_num) + '_step_k' + str(k) + '_d_' +
str(i) + '.txt', ant.resultsDir)
#hp.drawtree(pop[i].tree,ant.resultsDir + '_gen_' + str(gen) + '_pop_' + str(i))
f = open(
ant.resultsDir + 'lowlevel_gen_' + str(state.gen) +
'_pop_' + str(state.population_num) + '_step_k' + str(k) +
'_d_' + str(i) + '.txt', 'w')
#f.write(hp.tree2str(saved_tree[i]))
#f.write('#')
f.write('fitness: ' + str(m))
f.write('#')
if p:
f.write("_____exist best fitness")
if q:
f.write("___ that is better than overcome_desired")
#f.write('time: ' + str((time.time() - first_time)/60))
f.close()
###
fitness_k = min(y)
index = y.index(fitness_k)
if fitness_k < fitness:
for i in range(num):
current_chromosome[i] = X_training[index][i]
fitness = fitness_k
else:
anpha = anpha * lowinit.shrink # shrink the anpha maybe because the stepsize is a little big.
# np.savetxt('C:\\Opt_files\\lowlevel\\y.txt',y,delimiter = ',')
# np.savetxt('C:\\Opt_files\\lowlevel\\X.txt',X_training,delimiter = ',')
ind.fitness = fitness
ind.tree = hp.insert_chrom(ind.tree, current_chromosome, subtree_chrom,
IDlist)
state.lowlevel = False
#del saved_tree
return ind
def lowlevel_optimizer(ind,state,pop_num):
# the inputs: - IND individual
# - STATE
# - POP_NUM the specified ID of current individual which be optimized in here.
state.lowlevel = True
global_name = lowinit.tmpDir + ant.Antenna_Name + '_gen_'
global_tabname = lowinit.tmpTab + ant.Antenna_Name + '_gen_'
print('running low-level optimizer for generation ',state.gen,'population ', pop_num)
chromosome, subtree_chrom, IDlist = hp.getChrom(ind) # get chromosome from tree and other atributes.
num = len(chromosome) # number of values will be optimized.
# create randomly the trainning data for ANN.
y = np.zeros(lowinit.number_sample+1)
y[0] = ind.fitness # save original chromosome fitness.
return_loss = []
return_loss.append(ind.ReturnLoss) # save original return loss variable.
for i in range(lowinit.number_sample):
return_loss.append([])
X_training = np.zeros((lowinit.number_sample+1,num)) # X_train data
for i in range(num): # save original chromosome.
X_training[0,i] = chromosome[i]
for i in range(1,lowinit.number_sample+1):
for ii in range(num):
X_training[i][ii] = round(random.uniform(-1,1),4)
print(X_training)
# generate all of the scripts.
print('generating ',lowinit.number_sample, ' scripts in low lelvel optimizer in generation ',state.gen)
# before run the next generation, all of the tab file need be deleted.
filelist = [ f for f in os.listdir(lowinit.tmpDir) if f.endswith(".tab") ]
for f in filelist:
os.remove(os.path.join(lowinit.tmpDir, f))
# be fore run the next gen, all of the hfss file in temp need be deleted.
filelist = [ f for f in os.listdir(lowinit.tmpTab) if (f.endswith(".hfss") or f.endswith(".vbs") or f.endswith(".txt"))]
for f in filelist: # delete all before files.
os.remove(os.path.join(lowinit.tmpTab, f))
########################################################
############## insert chromosome.
# create temporary population to save current new individual.
#pop = []
#for i in range(1,lowinit.number_sample + 1):
# tree = hp.insert_chrom(ind.tree,X_training[i,:],subtree_chrom, IDlist)
for i in range(1,lowinit.number_sample + 1):
tree = hp.insert_chrom(ind.tree,X_training[i,:],subtree_chrom, IDlist)
#pop[i].tree.childs[1].valueofnode.plot()
[Substrate,polygons,centroid] = hp.get_all_para_for_hfss(tree) # get necessary parameters for genscript function.
name = global_name + str(state.gen) + '_pop_' + str(i) # name of directory would
# be used to save .vbs and .hfss file.
tabname = global_tabname + str(state.gen) + '_pop_' + str(i) # name of directory
# would be used to save .tab file.
temppp.tree = tree
temppp.nodelist = ind.nodelist
temp,_,_ = hp.getChrom(temppp)
print(X_training[i,:])
print(".....")
print(temp)
for iiii in range(len(X_training[i,:])):
if not (X_training[i,iiii] == temp[iiii]):
raise
genscript(Substrate,polygons,centroid,name + '.vbs',tabname,name + '.hfss')
del tree
del Substrate
del polygons
del centroid
#raise ValueError("Finished testing")
#######
# running all of the scripts.
nameDir = global_name + str(state.gen) + '_pop_' # file name direction of the vbs file.
hrun.RunPopScript(lowinit.number_sample,nameDir,1,len(init.PCnames),state)
# getting the fitness of each
for i in range(1,lowinit.number_sample+1):
spec_nameDir_tab = global_tabname + str(state.gen) + '_pop_' + str(i) + '.tab'
[m,n] = hrun.assignFitness(spec_nameDir_tab,state.gen)
print('i ',i, '___',m)
[y[i],return_loss[i]] = [m,n]
np.savetxt('C:\\Opt_files\\lowlevel\\y.txt',y,delimiter = ',')
np.savetxt('C:\\Opt_files\\lowlevel\\X.txt',X_training,delimiter = ',')
state.lowlevel = False