下面的代码运行，但它不起作用。变量cost总是一个充满1.0的Tensor，但为什么呢？我估计了一个标量，因为1x5矩阵乘以5x1矩阵是标量。在优化时偏差和权重也不会改变。我做错了什么？

#KI-Model
x = tf.placeholder(tf.float32, [None, 5], name='input') #x_1-x_5

#Init the graph
W = tf.Variable(tf.zeros([5,1]))
b = tf.Variable(tf.zeros([1]))

#activation with sigmoid
y = tf.nn.sigmoid(tf.matmul(x, W) + b) #berechneter Wert für y

#Training
y_tensor = tf.placeholder(tf.float32, [None, 1], name='output')
#cross_entropy = tf.reduce_mean(-tf.reduce_sum(y_tensor * tf.log(y), reduction_indices=[1])) #Hier Cross-Entropie statt minimum squares method

loss = y-y_tensor
cost = tf.square(loss)
optimizer = tf.train.GradientDescentOptimizer(0.001).minimize(cost)

#Start
session = tf.Session() #Google-> was ist das?
init = tf.global_variables_initializer()
session.run(init)

#init first 1000 training_batches
for i in range(1000):
    batch_xs.append([dataA[i], dataB[i], dataC[i], dataD[i],
              dataE[i]])
    batch_ys.append(dataG[i])

for i in range(10000):

    session.run(optimizer, feed_dict={x: batch_xs, y_tensor: batch_ys})

    print(session.run(cost, feed_dict={x: batch_xs, y_tensor: batch_ys}))

字符串