Tensorflow在评估我的测试模型时遇到问题

Question

下面是我的简单代码。目标是从数据中预测y=x²曲线。

import matplotlib.pyplot as plt
import tensorflow as tf
%matplotlib inline

#input data
x_data = np.linspace(0,10,11)
y_data=x_data**2

#Data for testing
x_test = np.linspace(0,10,11)

#placeholders
X = tf.placeholder("float32", [None, 11])
Y = tf.placeholder("float32", [None, 11])

#Neural Network Model
def multilayer_perceptron(X, weights, biases):
    layer_1 = tf.add(tf.matmul(X, weights['h1']), biases['b1'])
    layer_1 = tf.nn.sigmoid(layer_1)
    out_layer = tf.matmul(layer_1, weights['out']) + biases['out']
    return out_layer

# Store layers weight & bias
weights = {
    'h1': tf.Variable(tf.random_normal([11, 2], 0, 0.1)),
    'out': tf.Variable(tf.random_normal([2, 1], 0, 0.1))
}
biases = {
    'b1': tf.Variable(tf.random_normal([2], 0, 0.1)),
    'out': tf.Variable(tf.random_normal([1], 0, 0.1))
}

# Construct model
pred = multilayer_perceptron(X, weights, biases)

# Define loss and optimizer
cost = tf.reduce_mean(tf.square(pred-Y))
optimizer = tf.train.AdamOptimizer(learning_rate=0.001).minimize(cost)

init = tf.global_variables_initializer()

#Session Eval
with tf.Session() as sess:
    sess.run(init)
    epochs = 10000
    for i in range(epochs):
        o, c, p = sess.run([optimizer, cost, pred], feed_dict={X: x_data.reshape([1,11]),Y: y_data.reshape([1,11])})
    pred_value = sess.run([pred],feed_dict={X: x_test.reshape([1,11])})

最后一行是使用x_test作为输入调用一个会话来计算pred，对吗？当我运行这段代码时，pred_value只是一个随机数，而不是11个可能的正确值。我不知道代码出了什么问题

Answer 1

这是因为输出层的形状是(None, 1)。它应该是(None, 11)才能得到你想要的输出。因此，您必须将weights['out']的形状更改为(2,11)，将biases['out']的形状更改为(11)。

完整的工作代码如下。

import matplotlib.pyplot as plt
import tensorflow as tf
import numpy as np
# %matplotlib inline

#input data
x_data = np.linspace(0,10,11)
y_data=x_data**2

#Data for testing
x_test = np.linspace(0,10,11)

#placeholders
X = tf.placeholder("float32", [None, 11])
Y = tf.placeholder("float32", [None, 11])

#Neural Network Model
def multilayer_perceptron(X, weights, biases):
    layer_1 = tf.add(tf.matmul(X, weights['h1']), biases['b1'])
    layer_1 = tf.nn.sigmoid(layer_1)
    out_layer = tf.matmul(layer_1, weights['out']) + biases['out']
    print (out_layer.shape)
    return out_layer

# Store layers weight & bias
weights = {
    'h1': tf.Variable(tf.random_normal([11, 2], 0, 0.1)),
    'out': tf.Variable(tf.random_normal([2, 11], 0, 0.1))
}
biases = {
    'b1': tf.Variable(tf.random_normal([2], 0, 0.1)),
    'out': tf.Variable(tf.random_normal([11], 0, 0.1))
}

# Construct model
pred = multilayer_perceptron(X, weights, biases)

# Define loss and optimizer
cost = tf.reduce_mean(tf.square(pred-Y))
optimizer = tf.train.AdamOptimizer(learning_rate=0.001).minimize(cost)

init = tf.global_variables_initializer()

#Session Eval
with tf.Session() as sess:
    sess.run(init)
    epochs = 10000
    for i in range(epochs):
        o, c, p = sess.run([optimizer, cost, pred], feed_dict={X: x_data.reshape([1,11]),Y: y_data.reshape([1,11])})
        # print (p.shape)
    pred_value = sess.run([pred],feed_dict={X: x_test.reshape([1,11])})
    print (pred_value)

输出如下所示。

[array([[ 0.       ,  1.       ,  3.9999986,  8.999985 , 15.999548 ,
        23.496645 , 26.745462 , 28.187344 , 28.973213 , 28.841356 ,
        29.02704  ]], dtype=float32)]