Github:https://github.com/yjfiejd/Tensorflow_leaning/blob/master/tensorflow_20_RNN_lstm%20循环神经网络.py
# -*- coding:utf8 -*- # @TIME : 2018/4/30 上午11:34 # @Author : Allen # @File : RNN_lstm 循环神经网络.py #参考学习: # https://morvanzhou.github.io/tutorials/machine-learning/tensorflow/5-08-RNN2/ # Youtube video tutorial: https://www.youtube.com/channel/UCdyjiB5H8Pu7aDTNVXTTpcg # Youku video tutorial: http://i.youku.com/pythontutorial #这次我们会使用 RNN 来进行分类的训练 (Classification). 会继续使用到手写数字 MNIST 数据集. 让 RNN 从每张图片的第一行像素读到最后一行, 然后再进行分类判断. 接下来我们导入 MNIST 数据并确定 RNN 的各种参数(hyper-parameters): import tensorflow as tf from tensorflow.examples.tutorials.mnist import input_data #set random seed for comparing the two result calculations tf.set_random_seed(1) #导入数据 mnist = input_data.read_data_sets('MNIST_data', one_hot=True) #hyperparameters lr = 0.001 #学习率 training_iters = 100000 #循环次数上限 batch_size = 128 n_inputs = 28 #MNIST data input(img shape:28*28) 每一行28个像素点,28列 n_steps = 28 #time steps 每次input一行,一共28行 n_hidden_units = 128 #neurons in hidden layer n_classes = 10 #MNIST classes(0-9 digits) 把像书店分成10个类别 ####################################################### #接着定义 x, y 的 placeholder 和 weights, biases 的初始状况. #tf Graph input x = tf.placeholder(tf.float32, [None, n_steps, n_inputs]) #设置形参传递 y = tf.placeholder(tf.float32, [None, n_classes]) #Define weights, input -> intput_hidden_layer1 -> RNN cell -> output_hidden_layer2 -> output_layer #对weights biases 初始值定义 weights = { #(28, 128) 'in':tf.Variable(tf.random_normal([n_inputs, n_hidden_units])), #(128,10) 'out':tf.Variable(tf.random_normal([n_hidden_units, n_classes])) } biases = { #(128, ) 'in':tf.Variable(tf.constant(0.1, shape=[n_hidden_units, ])), #(10,) 'out':tf.Variable(tf.constant(0.1, shape=[n_classes,])) } # 定义RNN主体结构:共3个部分(input_layer, cell, output_layer) def RNN(X, weights, biases): #hidden layer for input to cell ############################### #原始对X是3维数据, 需要把它变为2维数据才能使用weights对矩阵乘法 #transpose the inputs shape from X, ==> (128*28, 28 inputs) X = tf.reshape(X, [-1, n_inputs]) #X_in = W*X + b X_in = tf.matmul(X, weights['in']) + biases['in'] # 格式重新变为3维数据 X_in ==> (128batch * 28steps, 128 hidden) X_in = tf.reshape(X_in, [-1, n_steps, n_hidden_units]) #cell 中对计算:使用 tf.nn.dynamic_rnn(cell, inputs) (推荐). 这次的练习将使用这种方式. ############################### #basic LSTM cell: lstm_cell = tf.contrib.rnn.BasicLSTMCell(n_hidden_units, forget_bias=1.0, state_is_tuple=True) #lstm cell is devide into two parts(c_state, m_state) 主线剧情,分线剧情 init_state = lstm_cell.zero_state(batch_size, dtype = tf.float32) #初始化全为0 state #如果使用tf.nn.dynamic_rnn(cell, inputs), 我们要确定 inputs 的格式. tf.nn.dynamic_rnn 中的 time_major 参数会针对不同 inputs 格式有不同的值. #如果 inputs 为 (batches, steps, inputs) ==> time_major=False; #如果 inputs 为 (steps, batches, inputs) ==> time_major=True; outputs, states = tf.nn.dynamic_rnn(lstm_cell, X_in, initial_state=init_state, time_major=False) #hidden layer for output as the final results ################################## #方法1:直接调用final_state 中的 h_state (final_state[1]) 来进行运算: results = tf.matmul(states[1], weights['out']) + biases['out'] #方法2: #outputs = tf.unpack(tf.transpose(outputs, [1, 0, 2])) #results = tf.matmul(outputs[-1], weights['out'] + biases['out']) return results #定义好了 RNN 主体结构后, 我们就可以来计算 cost 和 train_op: pred = RNN(x, weights, biases) cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits = pred,labels = y)) train_op = tf.train.AdamOptimizer(lr).minimize(cost) # 训练RNN correct_pred = tf.equal(tf.argmax(pred, 1), tf.argmax(y, 1)) accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32)) init = tf.global_variables_initializer() with tf.Session() as sess: sess.run(init) step = 0 while step*batch_size < training_iters: batch_xs, batch_ys = mnist.train.next_batch(batch_size) batch_xs = batch_xs.reshape([batch_size, n_steps, n_inputs]) sess.run([train_op], feed_dict={x:batch_xs, y:batch_ys}) if step % 20 == 0: print(sess.run(accuracy, feed_dict={x:batch_xs, y:batch_ys})) step += 1 #结果如下: #2018-04-30 14:05:50.942937: I tensorflow/core/platform/cpu_feature_guard.cc:140] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2 FMA #0.265625 #0.7265625 #0.828125 #0.8828125 #0.84375 #0.859375 #0.8984375 #0.890625 #0.84375 #0.90625 #0.921875 #0.90625 #0.9140625 #0.9140625 #0.9375 #0.9609375 #0.953125 #0.921875 #0.9453125 #0.96875 #0.9375 #0.9609375