テンソルフローにおけるMNISTデータベースの大出力予測

Question

テスト例でネットワークのトレーニング後に結果を受け取ることができません。 それは

examples_to_show = 5 
y_result = sess.run(y_pred, feed_dict={x:mnist.test.images[:examples_to_show]}) 
print("y_result=",y_result) 

私が代わりに[0 0 1 0 0 0 0 0 0 0のない不明瞭な数字を受け取る

は、私は、このような方法で結果を受信しようヘルプmultilayer_perceptron.pyから標準の例です]

In [20]: 
''' 
A Multilayer Perceptron implementation example using TensorFlow library. 
This example is using the MNIST database of handwritten digits 
Author: Aymeric Damien 
''' 

In [21]: 
# Import MINST data 
from tensorflow.examples.tutorials.mnist import input_data 
mnist = input_data.read_data_sets("MNIST_data/", one_hot=True) 
import tensorflow as tf 

Extracting MNIST_data/train-images-idx3-ubyte.gz 
Extracting MNIST_data/train-labels-idx1-ubyte.gz 
Extracting MNIST_data/t10k-images-idx3-ubyte.gz 
Extracting MNIST_data/t10k-labels-idx1-ubyte.gz 

In [22]: 
# Parameters 
learning_rate = 0.001 
training_epochs = 15 
batch_size = 100 
display_step = 1 
# Network Parameters 
n_hidden_1 = 256 # 1st layer number of features 
n_hidden_2 = 256 # 2nd layer number of features 
n_input = 784 # MNIST data input (img shape: 28*28) 
n_classes = 10 # MNIST total classes (0-9 digits) 
# tf Graph input 
x = tf.placeholder("float", [None, n_input]) 
y = tf.placeholder("float", [None, n_classes]) 

In [23]: 
# Create model 
def multilayer_perceptron(x, weights, biases): 
# Hidden layer with RELU activation 
layer_1 = tf.add(tf.matmul(x, weights['h1']), biases['b1']) 
layer_1 = tf.nn.relu(layer_1) 
# Hidden layer with RELU activation 
layer_2 = tf.add(tf.matmul(layer_1, weights['h2']), biases['b2']) 
layer_2 = tf.nn.relu(layer_2) 
# Output layer with linear activation 
out_layer = tf.matmul(layer_2, weights['out']) + biases['out'] 
return out_layer 

In [24]: 
# Store layers weight & bias 
weights = { 
    'h1': tf.Variable(tf.random_normal([n_input, n_hidden_1])), 
    'h2': tf.Variable(tf.random_normal([n_hidden_1, n_hidden_2])), 
    'out': tf.Variable(tf.random_normal([n_hidden_2, n_classes])) 
} 
biases = { 
    'b1': tf.Variable(tf.random_normal([n_hidden_1])), 
    'b2': tf.Variable(tf.random_normal([n_hidden_2])), 
    'out': tf.Variable(tf.random_normal([n_classes])) 
} 

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

# Define loss and optimizer 
cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=pred, labels=y)) 
optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate).minimize(cost) 

# Initializing the variables 
init = tf.global_variables_initializer() 

The designer for predictions!!! 
In [25]: 

# Prediction 
y_pred = pred 

In [30]: 
# Launch the graph 
with tf.Session() as sess: 
    sess.run(init) 
    # Training cycle 
    for epoch in range(training_epochs): 
     avg_cost = 0. 
     total_batch = int(mnist.train.num_examples/batch_size) 
     # Loop over all batches 
     for i in range(total_batch): 
     batch_x, batch_y = mnist.train.next_batch(batch_size) 
     # Run optimization op (backprop) and cost op (to get loss value) 
     _, c = sess.run([optimizer, cost], feed_dict={x: batch_x, 
                 y: batch_y}) 
     # Compute average loss 
     avg_cost += c/total_batch 
    # Display logs per epoch step 
    if epoch % display_step == 0: 
     print("Epoch:", '%04d' % (epoch+1), "cost=", \ 
      "{:.9f}".format(avg_cost)) 
    print("Optimization Finished!") 

    # Test model 
    correct_prediction = tf.equal(tf.argmax(pred, 1), tf.argmax(y, 1)) 
    # Calculate accuracy 
    accuracy = tf.reduce_mean(tf.cast(correct_prediction, "float")) 
    print("Accuracy:", accuracy.eval({x: mnist.test.images, y: mnist.test.labels})) 

    # We will try to receive result of training!!! 
    examples_to_show = 5 
    y_result = sess.run(y_pred, feed_dict={x: mnist.test.images[:examples_to_show]}) 
    print("y_result=",y_result) 


Epoch: 0001 cost= 142.664078834 
Epoch: 0002 cost= 37.176684845 
Epoch: 0003 cost= 23.608409217 
Epoch: 0004 cost= 16.678811304 
Epoch: 0005 cost= 12.175642554 
Epoch: 0006 cost= 9.083989911 
Epoch: 0007 cost= 6.624555320 
Epoch: 0008 cost= 4.970751049 
Epoch: 0009 cost= 3.595181121 
Epoch: 0010 cost= 2.671157273 
Epoch: 0011 cost= 2.032964239 
Epoch: 0012 cost= 1.588672840 
Epoch: 0013 cost= 1.133152580 
Epoch: 0014 cost= 0.805134769 
Epoch: 0015 cost= 0.689760053 
Optimization Finished! 
Accuracy: 0.941 

y_result= [ 
[ -203.50767517 -437.82525635 186.90861511 590.15588379 
-471.18536377 -283.88424683 -1150.14709473 1022.75799561 -391.6159668 
432.9206543 ] 
[ -855.87487793  6.88715792 903.70776367 252.00227356 
-1407.09313965 441.29104614 344.09405518 -1691.98535156 
40.62039566 -1391.43688965] 
[ -244.32698059 618.91705322 12.79210854 -36.14464951 
-8.12554073 183.12348938 50.32661057 147.05378723 152.9332428 
-210.40829468] 
[ 1091.7199707 -919.26574707 -333.54571533 -953.7399292 -1072.82226562 
73.99294281 305.2588501 -166.91053772 -985.14654541 
452.14318848] 
[ 200.62698364 89.34638214 -280.01904297 -342.19534302 1240.4128418 
229.24633789 -424.91091919 298.81100464 -194.70623779 
934.27703857]] 

結果はy_result = [0 0 1 0 0 0 0 0 0 0]でなければならない！ ??? ???なぜ ？？？

Answer 1

あなたのpredには、ロジットを確率に変換する出力アクティベーションがありません。したがって、tf.softmax(pred)を適用し、これをあなたの予測として使用してください。 softmaxを内部的に適用するので、これをsoftmax_cross_entropy()に渡さないように注意してください。 コードを次のように変更できます。

# Construct model 
logits = multilayer_perceptron(x, weights, biases) 
cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=logits, labels=y)) 

# Apply softmax to obtain probabilities 
pred = tf.nn.softmax(logits) 

Answer 2

y_resultは、out_layer = tf.matmul(layer_2, weights['out']) + biases['out']です。 1つのベクトルではなく、行列またはベクトル（layer_2およびweight['out']に応じて）のいずれかになります。結果を見るのはマトリックスです