def gridSearch(X,
instance,
numAnts,
max_evals,
max_gens,
rho_list,
alpha_list,
beta_list,
ph_min_list,
ph_max_list,
BBO=True):
plt.ion()
fig, ax = plt.subplots()
ax.set_xlabel("# evaluations")
ax.set_ylabel("fitness")
lab = ""
# cm = plt.get_cmap('gist_rainbow')
cm = plt.get_cmap('tab20')
markers = ["x", "^"]
markerind = 0
NUM_COLORS = len(rho_list) * len(alpha_list) * len(beta_list)
ax.set_prop_cycle(
color=[cm(1. * i / NUM_COLORS) for i in range(NUM_COLORS)])
for rho in rho_list:
for alpha in alpha_list:
for beta in beta_list:
for ph_min in ph_min_list:
for ph_max in ph_max_list:
if BBO:
ACO_instance = ACO.ACO_BBO2(
instance, numAnts, max_evals, max_gens, rho,
ph_max, ph_min, alpha)
lab = r"$\rho = %.2f, \alpha = %.2f $" % (rho,
alpha)
else:
ACO_instance = ACO.ACO_GBO(instance, numAnts,
max_evals, max_gens,
rho, ph_max, ph_min,
alpha, beta)
lab = r"$\rho = %.2f, \alpha = %.2f, \beta = %.2f $" % (
rho, alpha, beta)
x, y, e, _ = rn.executeXtimes(X, ACO_instance)
ax.errorbar(x,
y,
e,
linestyle='None',
marker=markers[markerind],
label=lab)
ax.legend(loc="lower right")
plt.draw(), plt.pause(1e-4)
markerind = 1 - markerind
plt.ioff()
def doGlobalTest():
np.set_printoptions(formatter={'float': lambda x: "{0:0.2f}".format(x)})
instances_directory = 'testinstances/'
opt_directory = 'opts/'
instancenamelist = [
"maxcut_random_16_0.4_100_instance_0",
"maxcut_random_32_0.4_100_instance_1",
"maxcut_random_64_0.4_100_instance_2",
"maxcut_random_128_0.4_100_instance_3",
"maxcut_random_256_0.4_100_instance_4",
"maxcut_random_512_0.4_100_instance_5"
]
# instancenamelist = ["maxcut_random_16_0.4_100_instance_0","maxcut_random_32_0.4_100_instance_1"]
plt.ion()
f = open("ACOend_results.txt", "a")
f.write(
"filename \t\t\t\t\t\t\t evaluations \t mean \t std \t maxfitness \n")
# f2 = open("ACOrun_results.txt","w")
for instancename in instancenamelist:
fig, ax = plt.subplots()
instance = maxcut.MaxCut(instancename + ".txt", instances_directory,
opt_directory)
numAnts = 20
max_evals = 10000000000
max_gens = 200
# best parameters for instance newL50_1_opt2078 without local search, rho = 0.2
rho = 0.3 # result for with local search on N=128, 0.3 or 0.35
ph_max = 100
ph_min = 0.1
alpha = 0.7
ACO_instance = ACO.ACO_BBO2(instance, numAnts, max_evals, max_gens,
rho, ph_max, ph_min, alpha)
X = 10
x, y, e, maxfit = rn.executeXtimes(X, ACO_instance)
ax.errorbar(x, y, e, linestyle='None', marker="x", label=instancename)
ax.legend(loc="lower right")
plt.draw(), plt.pause(1e-4)
f.write("%s \t %.0f \t %.2f \t %.5f \t %.0f \n" %
(instancename, x[-1], y[-1], e[-1], maxfit))
f.flush()
# f.write("end mean: %.2f \n" % y[-1])
# f.write("end std: %.5f \n" % e[-1])
# f.write("best fitness found: %.0f" % maxfit)
plt.ioff()
f.close()
plt.show()
max_gens = 100
# rho = 1 - 0.01**(1/(max_evals/4))
#
rho = 0.21
ph_max=10
ph_min=0.1
alpha = 0.7
fig, ax = plt.subplots()
fig2, ax2 = plt.subplots()
ACO_BBO = ACO.ACO_BBO(instance,numAnts,max_evals,max_gens,rho,ph_max,ph_min,alpha,ax)
ACO_BBO.run()
ACO_BBO2 = ACO.ACO_BBO2(instance,numAnts,max_evals,max_gens,rho,ph_max,ph_min,alpha,ax)
ACO_BBO2.run()
# GBO
rho = 0.41
alpha = 1.5
beta = 1
# ACO_GBO = ACO.ACO_GBO(instance,numAnts,max_evals,max_gens,rho,ph_max,ph_min,alpha,beta,ax)
# ACO_GBO.run()
# ACO_GBO2 = ACO.ACO_GBO2(instance,numAnts,max_evals,max_gens,rho,ph_max,ph_min,alpha,beta,ax)
# ACO_GBO2.run()
ax2.plot(ACO_BBO.numEvalsList,ACO_BBO.archiveElitistList,color="black",label="Black Box archive elitist fitness")
ax2.plot(ACO_BBO.numEvalsList,ACO_BBO.averageFitnessList,'-*',color="black",label="Black Box average population fitness")