单层LSTM网络对MNIST数据集分类

单层LSTM网络对MNIST数据集分类

实验代码：（使用tensorflow框架）

# -*- coding: utf-8 -*-

import tensorflow as tf
# 导入 MINST 数据集
from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets("/data/", one_hot=True)

n_input = 28 # MNIST data 输入 (img shape: 28*28)
n_steps = 28 # timesteps
n_hidden = 128 # hidden layer num of features
n_classes = 10  # MNIST 列别 (0-9 ，一共10类)

tf.reset_default_graph()

# tf Graph input
x = tf.placeholder("float", [None, n_steps, n_input])
y = tf.placeholder("float", [None, n_classes])


x1 = tf.unstack(x, n_steps, 1)

#1 BasicLSTMCell
lstm_cell = tf.contrib.rnn.BasicLSTMCell(n_hidden, forget_bias=1.0)
outputs, states = tf.contrib.rnn.static_rnn(lstm_cell, x1, dtype=tf.float32)

#2 LSTMCell
#lstm_cell = tf.contrib.rnn.LSTMCell(n_hidden, forget_bias=1.0)
#outputs, states = tf.contrib.rnn.static_rnn(lstm_cell, x1, dtype=tf.float32)

#3 gru
#gru = tf.contrib.rnn.GRUCell(n_hidden)
#outputs = tf.contrib.rnn.static_rnn(gru, x1, dtype=tf.float32)

#4 创建动态RNN
#outputs,_  = tf.nn.dynamic_rnn(gru,x,dtype=tf.float32)
#outputs = tf.transpose(outputs, [1, 0, 2])

pred = tf.contrib.layers.fully_connected(outputs[-1],n_classes,activation_fn = None)



learning_rate = 0.001
training_iters = 100000
batch_size = 128
display_step = 10

# 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)

# Evaluate model
correct_pred = tf.equal(tf.argmax(pred,1), tf.argmax(y,1))
accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))

# 启动session
with tf.Session() as sess:
    sess.run(tf.global_variables_initializer())
    step = 1
    # Keep training until reach max iterations
    while step * batch_size < training_iters:
        batch_x, batch_y = mnist.train.next_batch(batch_size)
        # Reshape data to get 28 seq of 28 elements
        batch_x = batch_x.reshape((batch_size, n_steps, n_input))
        # Run optimization op (backprop)
        sess.run(optimizer, feed_dict={x: batch_x, y: batch_y})
        if step % display_step == 0:
            # 计算批次数据的准确率
            acc = sess.run(accuracy, feed_dict={x: batch_x, y: batch_y})
            # Calculate batch loss
            loss = sess.run(cost, feed_dict={x: batch_x, y: batch_y})
            print ("Iter " + str(step*batch_size) + ", Minibatch Loss= " + \
                  "{:.6f}".format(loss) + ", Training Accuracy= " + \
                  "{:.5f}".format(acc))
        step += 1
    print (" Finished!")

    # 计算准确率 for 128 mnist test images
    test_len = 128
    test_data = mnist.test.images[:test_len].reshape((-1, n_steps, n_input))
    test_label = mnist.test.labels[:test_len]
    print ("Testing Accuracy:", \
        sess.run(accuracy, feed_dict={x: test_data, y: test_label}))

实验结果：
Iter 1280, Minibatch Loss= 2.098885, Training Accuracy= 0.30469
Iter 2560, Minibatch Loss= 1.772232, Training Accuracy= 0.38281
Iter 3840, Minibatch Loss= 1.404505, Training Accuracy= 0.52344
Iter 5120, Minibatch Loss= 1.321466, Training Accuracy= 0.57031
Iter 6400, Minibatch Loss= 1.020606, Training Accuracy= 0.65625
Iter 7680, Minibatch Loss= 0.767583, Training Accuracy= 0.76562
Iter 8960, Minibatch Loss= 0.945606, Training Accuracy= 0.66406
Iter 10240, Minibatch Loss= 0.643211, Training Accuracy= 0.78906
Iter 11520, Minibatch Loss= 0.737389, Training Accuracy= 0.76562
Iter 12800, Minibatch Loss= 0.589967, Training Accuracy= 0.83594
Iter 14080, Minibatch Loss= 0.432091, Training Accuracy= 0.89062
Iter 15360, Minibatch Loss= 0.375092, Training Accuracy= 0.90625
Iter 16640, Minibatch Loss= 0.509971, Training Accuracy= 0.82031
Iter 17920, Minibatch Loss= 0.431015, Training Accuracy= 0.85156
Iter 19200, Minibatch Loss= 0.420453, Training Accuracy= 0.85156
Iter 20480, Minibatch Loss= 0.338827, Training Accuracy= 0.88281
Iter 21760, Minibatch Loss= 0.427024, Training Accuracy= 0.86719
Iter 23040, Minibatch Loss= 0.419629, Training Accuracy= 0.87500
Iter 24320, Minibatch Loss= 0.343750, Training Accuracy= 0.90625
Iter 25600, Minibatch Loss= 0.232130, Training Accuracy= 0.92188
Iter 26880, Minibatch Loss= 0.491618, Training Accuracy= 0.89062
Iter 28160, Minibatch Loss= 0.226970, Training Accuracy= 0.92188
Iter 29440, Minibatch Loss= 0.287028, Training Accuracy= 0.91406
Iter 30720, Minibatch Loss= 0.348053, Training Accuracy= 0.90625
Iter 32000, Minibatch Loss= 0.232494, Training Accuracy= 0.92969
Iter 33280, Minibatch Loss= 0.294077, Training Accuracy= 0.89062
Iter 34560, Minibatch Loss= 0.269400, Training Accuracy= 0.90625
Iter 35840, Minibatch Loss= 0.257503, Training Accuracy= 0.92969
Iter 37120, Minibatch Loss= 0.176288, Training Accuracy= 0.95312
Iter 38400, Minibatch Loss= 0.263634, Training Accuracy= 0.89844
Iter 39680, Minibatch Loss= 0.350406, Training Accuracy= 0.89062
Iter 40960, Minibatch Loss= 0.175449, Training Accuracy= 0.94531
Iter 42240, Minibatch Loss= 0.311644, Training Accuracy= 0.89844
Iter 43520, Minibatch Loss= 0.202412, Training Accuracy= 0.92188
Iter 44800, Minibatch Loss= 0.238732, Training Accuracy= 0.92188
Iter 46080, Minibatch Loss= 0.262362, Training Accuracy= 0.91406
Iter 47360, Minibatch Loss= 0.277031, Training Accuracy= 0.92188
Iter 48640, Minibatch Loss= 0.167007, Training Accuracy= 0.93750
Iter 49920, Minibatch Loss= 0.208343, Training Accuracy= 0.95312
Iter 51200, Minibatch Loss= 0.237634, Training Accuracy= 0.91406
Iter 52480, Minibatch Loss= 0.133993, Training Accuracy= 0.96094
Iter 53760, Minibatch Loss= 0.255377, Training Accuracy= 0.92188
Iter 55040, Minibatch Loss= 0.204812, Training Accuracy= 0.92969
Iter 56320, Minibatch Loss= 0.183624, Training Accuracy= 0.92969
Iter 57600, Minibatch Loss= 0.131443, Training Accuracy= 0.96094
Iter 58880, Minibatch Loss= 0.096448, Training Accuracy= 0.97656
Iter 60160, Minibatch Loss= 0.163977, Training Accuracy= 0.96875
Iter 61440, Minibatch Loss= 0.185323, Training Accuracy= 0.95312
Iter 62720, Minibatch Loss= 0.107512, Training Accuracy= 0.97656
Iter 64000, Minibatch Loss= 0.174152, Training Accuracy= 0.95312
Iter 65280, Minibatch Loss= 0.173235, Training Accuracy= 0.95312
Iter 66560, Minibatch Loss= 0.115825, Training Accuracy= 0.96875
Iter 67840, Minibatch Loss= 0.190322, Training Accuracy= 0.92969
Iter 69120, Minibatch Loss= 0.073072, Training Accuracy= 0.97656
Iter 70400, Minibatch Loss= 0.161416, Training Accuracy= 0.93750
Iter 71680, Minibatch Loss= 0.148715, Training Accuracy= 0.95312
Iter 72960, Minibatch Loss= 0.174622, Training Accuracy= 0.95312
Iter 74240, Minibatch Loss= 0.100780, Training Accuracy= 0.97656
Iter 75520, Minibatch Loss= 0.177840, Training Accuracy= 0.96094
Iter 76800, Minibatch Loss= 0.119568, Training Accuracy= 0.96094
Iter 78080, Minibatch Loss= 0.116565, Training Accuracy= 0.96094
Iter 79360, Minibatch Loss= 0.124705, Training Accuracy= 0.96094
Iter 80640, Minibatch Loss= 0.068246, Training Accuracy= 0.97656
Iter 81920, Minibatch Loss= 0.152009, Training Accuracy= 0.97656
Iter 83200, Minibatch Loss= 0.150834, Training Accuracy= 0.96094
Iter 84480, Minibatch Loss= 0.082806, Training Accuracy= 0.98438
Iter 85760, Minibatch Loss= 0.239210, Training Accuracy= 0.94531
Iter 87040, Minibatch Loss= 0.194339, Training Accuracy= 0.94531
Iter 88320, Minibatch Loss= 0.141747, Training Accuracy= 0.96094
Iter 89600, Minibatch Loss= 0.110870, Training Accuracy= 0.97656
Iter 90880, Minibatch Loss= 0.066232, Training Accuracy= 0.98438
Iter 92160, Minibatch Loss= 0.085497, Training Accuracy= 0.96875
Iter 93440, Minibatch Loss= 0.141791, Training Accuracy= 0.96094
Iter 94720, Minibatch Loss= 0.143089, Training Accuracy= 0.93750
Iter 96000, Minibatch Loss= 0.234196, Training Accuracy= 0.93750
Iter 97280, Minibatch Loss= 0.143507, Training Accuracy= 0.94531
Iter 98560, Minibatch Loss= 0.069923, Training Accuracy= 0.96875
Iter 99840, Minibatch Loss= 0.079662, Training Accuracy= 0.98438
Finished!
Testing Accuracy: 0.976562

---

参考资料：《深度学习之Tensorflow》李金洪编著