# import tensorflow as tf import numpy as np from sa import SA from common import * model = SA(shape=[48, 48, 1], no_classes=7) model.build_model() data = np.zeros((7, 100, 48, 48, 1)) val = np.ones((20, 48, 48, 1)) labels_val = one_hot(np.ones((20)), 7) model.train(data, val, labels_val, period_save_model=5)
# -*- coding: utf-8 -*-
"""
experiment on different anneal rate(gamma). For more detail, please reference to Readme.md
"""
import numpy as np
from sa import SA
data = np.loadtxt('location.txt', delimiter=',', usecols=[1, 2])
seed = None
sa = SA(data,
T=200,
anneal_rate=0.95,
final_T=0.001,
inner_iters=120,
random_init=True)
gammas = [0.1, 0.3, 0.5, 0.8, 0.9, 0.95, 0.98, 0.99]
n = 5
with open('./experiment/2.3_gamma.txt', 'w') as f:
for gamma in gammas:
t = 0
dis = 0
for i in range(n):
sa.reset()
sa.gamma = gamma
log = sa.train(mode='reverse')
t += log['time']
dis += log['dis']
s = 'gamma={}\ttime={:.3f}\tdistance={}'.format(gamma, t / n, dis / n)
print(s, file=f, flush=True)
print(s)
type=int,
default=0,
help='anneal mode. It must be one of [0,1,2](default:0)')
ars = parser.parse_args()
np.random.seed(ars.seed) # set random seed for reproducing experiment results
data = np.loadtxt(ars.file_path, delimiter=',', usecols=[1, 2]) # load data
# init algorithm
sa = SA(data,
T=ars.init_T,
anneal_rate=ars.anneal_rate,
final_T=ars.final_T,
inner_iters=ars.inner_iters,
random_init=ars.random_init)
# run algorithm
log = sa.train(mode=ars.mode, anneal_mode=ars.anneal_mode)
# log information
t = log['time']
dis = log['dis']
path = log['path']
dis_lst = deepcopy(sa.dis_lst)
print('time:{:.3f}\tbest distance:{}'.format(t, dis))
print('Now you can use test.exe to verify solution.')
# save best result. So you can excute test.exe to verify the solution
np.savetxt('city.txt', path + 1, fmt='%d', newline=',')
# plot distance over time
fig = plt.figure()
ax = fig.add_subplot(211)
ax.plot(dis_lst)
ax.set_xlabel('num')
ax.set_ylabel('distance')
# -*- coding: utf-8 -*-
"""
experiment on different state transition mode. For more detail, please reference to Readme.md
"""
import numpy as np
from sa import SA
data = np.loadtxt('location.txt', delimiter=',', usecols=[1, 2])
seed = None
sa = SA(data,
T=200,
anneal_rate=0.95,
final_T=0.001,
inner_iters=120,
random_init=True)
modes = ['cross', 'reverse']
n = 10
with open('./experiment/2.5_mode.txt', 'w') as f:
for mode in modes:
t = 0
dis = 0
for i in range(n):
sa.reset()
log = sa.train(mode=mode)
t += log['time']
dis += log['dis']
s = 'mode={}\ttime={:.3f}\tdistance={}'.format(mode, t / n, dis / n)
print(s, file=f, flush=True)
print(s)
import numpy as np
from sa import SA
from copy import deepcopy
data = np.loadtxt('location.txt', delimiter=',', usecols=[1, 2])
seed = None
sa = SA(data,
T=200,
anneal_rate=0.95,
final_T=0.001,
inner_iters=120,
random_init=True)
modes = [0, 1, 2]
n = 5
dis_lst = None
with open('./experiment/2.2_annealMode.txt', 'w') as f:
for mode in modes:
t = 0
dis = 0
for i in range(n):
sa.reset()
log = sa.train(mode='reverse', anneal_mode=mode)
t += log['time']
dis += log['dis']
dis_lst = deepcopy(sa.dis_lst)
sa.plot_dis(title=f'Mode {mode}')
s = 'anneal mode={}\ttime={:.3f}\tdistance={}'.format(
mode, t / n, dis / n)
print(s, file=f, flush=True)
print(s)