def min_goal(x):
m = []
for i in range(len(encoder_result)):
m.append(function.func1(x[i]))
#print(m.index(min(m)))
return m.index(min(m))
def min_goal(self, x):
m = []
for i in range(len(x)):
m.append(function.func1(x[i]))
#print(m.index(min(m)))
return m.index(min(m))
import function #This imports all the function in the function python file import random function.func1() x = random.randrange(1, 1100) print(x)
def run(self):
time_start = time.time()
sess = tf.Session()
encoder_op = self.encoder(self.array) # 编码
decoder_op = self.decoder(encoder_op) # 解码
y_pred = decoder_op
y_true = self.X
## 定义代价函数和优化器
# 最小二乘法 rmse
cost = tf.reduce_mean(tf.pow(y_true - y_pred, 2))
rmse = cost**0.5
# 优化函数 全局学习速率 0.01
optimizer = tf.train.AdamOptimizer(1).minimize(rmse)
if int((tf.__version__).split('.')[1]) < 12 and int(
(tf.__version__).split('.')[0]) < 1:
init = tf.initialize_all_variables()
else:
init = tf.global_variables_initializer()
sess.run(init)
encoder_result = sess.run(encoder_op, feed_dict={self.X: self.array})
total_batch = int(50 / self.batch_size) # 总批数55000/256
for epoch in range(self.training_epochs): # 训练次数
for i in range(total_batch): # 每次训练遍历训练集所需次数
#batch_xs, batch_ys = tf.train.batch([y_pred, y_true], batch_size=batch_size)
batch = self.get_random_block_from_data(
self.array, self.batch_size)
_, c = sess.run([optimizer, rmse], feed_dict={self.X: batch})
#print(c)
print("Optimization Finished!")
X = []
Y = []
Z = []
# 读入n个点的坐标
for i in range(6):
a = random.randint(0, 30)
X.append(encoder_result[a][0])
Y.append(encoder_result[a][1])
for i in range(len(encoder_result)):
Z.append(function.func1(encoder_result[a]))
def f(x):
a, b, c, d, e, f = self.FindFunction(6, X, Y, Z)
x2 = x[1]
x1 = x[0]
return [2 * a * x1 + b * x2 + d, b * x1 + 2 * c * x2 + e]
result = fsolve(f, [1, 1])
#print(result)
decoder_result = sess.run(decoder_op, feed_dict={self.X: self.array})
# print('维度还原之后的输出')
#print(decoder_result)
encoder_result[-1] = result
# print(encoder_result)
a = []
for i in range(len(encoder_result)):
a.append(encoder_result[i][0])
b = []
for i in range(len(encoder_result)):
b.append(encoder_result[i][1])
plt.scatter(a, b)
#print('位置')
#print(self.min_goal(encoder_result))
#最终得出的粒子 decoder_result[min_goal(encoder_result)]
#print('30维结果')
#print(decoder_result[self.min_goal(encoder_result)])
# a = random.choice(self.array)
#print(decoder_result[-1])
a = random.randint(0, 30)
del self.array[a]
e = decoder_result[-1].tolist()
print(e)
m = function.func1(e)
print(m)
self.array.append(e)
#print('最后的列表')
#print(self.array)
#print(type(e))
#plt.show()
time_end = time.time()
#print('time cost_autoencoder', time_end - time_start, 's')
tf.reset_default_graph()
return self.array