Tensorflow 学习笔记

摘要

研二可能搞大数据，Spark和数据挖掘、机器学习什么的可能都要用到，做个笔记，顺带分享

Reference
[1]. http://wiki.jikexueyuan.com/project/tensorflow-zh/tutorials/deep_cnn.html
[2]. http://v.youku.com/v_show/id_XMjc3MzA3OTM0NA==.html?spm=a2hzp.8253876.0.0&f=27327189

基于Win10 Tensorflow 0.12(CPU)

简单的实例

(1) 一个最简单的直线方程

import tensorflow as tfimport numpy as np# create dataprint('create data...')x_data = np.random.rand(100).astype(np.float32)y_data = x_data * 0.1 + 0.3# create tensorflow structure startWeights = tf.Variable(tf.random_uniform([1],-1.0,1.0))biases = tf.Variable(tf.zeros([1]))y = Weights*x_data + biasesloss = tf.reduce_mean(tf.square(y-y_data))learn_effciency = 0.5 #学习效率一般小于1optimizer = tf.train.GradientDescentOptimizer(learn_effciency)train = optimizer.minimize(loss)init = tf.initialize_all_variables() #初始化所有变量# Start the neutral networksess = tf.Session()sess.run(init)for step in range(201):    sess.run(train)    if step%20==0 :        print(step,sess.run(Weights),sess.run(biases))

(2) Session的两种使用方式

import tensorflow as tfimport numpy as npmatrix1 = tf.constant([[3,3]])matrix2 = tf.constant([[2],[2]])product = tf.matmul(matrix1,matrix2) # matrix multiply == np.dot(m1,m2)# # Session usage type 1# sess = tf.Session()# result = sess.run(product)# print(result)# sess.close()# Session usage type 2with tf.Session() as sess: # 类似于一个auto_ptr    result2 = sess.run(product)    print(result2)

(3) 使用变量一定要记得run初始化哦

import tensorflow as tfstate = tf.Variable(0, name='counter')one = tf.constant(1)new_value = tf.add(state,one)update = tf.assign(state, new_value)# 必须初始化所有变量# Deprecated and will be removed after 2017-03-02.# init = tf.initialize_all_variables() # with tf.Session() as sess:#     sess.run(init)#     for _ in range(3):#         sess.run(update)#         print(sess.run(state))# Latest solutioninit_op = tf.global_variables_initializer()local_init_op = tf.local_variables_initializer()  # local variables like epoch_num, batch_size 可以不初始化localwith tf.Session() as sess:    sess.run(init_op)    sess.run(local_init_op)    for _ in range(3):        sess.run(update)        print(sess.run(state))

(4) 使用placeholder预定义变量

import tensorflow as tfinput1 = tf.placeholder(tf.float32)input2 = tf.placeholder(tf.float32)output = tf.multiply(input1,input2) # tf.mul已经用不了了with tf.Session() as sess:    print(sess.run(output,feed_dict={input1:[7.], input2:[2.]})) # 补上数据

(5) 添加神经层

import tensorflow as tf# 添加神经层def add_layer(inputs, in_size, out_size, activation_function=None):    Weights = tf.Variable(tf.random_normal([in_size, out_size]))    biases = tf.Variable(tf.zeros([1, out_size]) + 0.1)    Wx_plus_b = tf.matmul(inputs, Weights) + biases # 没有被激活的值    if activation_function is None: # 说明是线性方程        outputs = Wx_plus_b    else:        outputs = activation_function(Wx_plus_b)    return outputs

(6) 一个只有一层hidden layer的简单学习

import tensorflow as tfimport numpy as np# 添加神经层def add_layer(inputs, in_size, out_size, activation_function=None):    Weights = tf.Variable(tf.random_normal([in_size, out_size]))    biases = tf.Variable(tf.zeros([1, out_size]) + 0.1)    Wx_plus_b = tf.matmul(inputs, Weights) + biases # 没有被激活的值    if activation_function is None: # 说明是线性方程        outputs = Wx_plus_b    else:        outputs = activation_function(Wx_plus_b)    return outputsx_data = np.linspace(-1,1,300)[:,np.newaxis] # 后面的中括号用于加维度noise = np.random.normal(0,0.05,x_data.shape) # 噪声，用来模拟真实数据y_data = np.square(x_data) - 0.5 + noisexs = tf.placeholder(tf.float32, [None, 1]) # None表示无论给多少个sample都可以ys = tf.placeholder(tf.float32, [None, 1])# 1 -> 10 -> 1l1 = add_layer(xs, 1, 10, activation_function=tf.nn.relu)prediction = add_layer(l1, 10, 1, activation_function=None)loss = tf.reduce_mean(   # 求平均数          tf.reduce_sum( # 对例子的差进行求和            tf.square(ys - prediction), reduction_indices=[1]))train_step = tf.train.GradientDescentOptimizer(0.1).minimize(loss)init = tf.global_variables_initializer()with tf.Session() as sess:    sess.run(init)    for i in range(1000):        sess.run(train_step,feed_dict={xs:x_data, ys:y_data})        if i% 50==0:            print(sess.run(loss,feed_dict={xs:x_data, ys:y_data}))

(7) 训练过程可视化

import tensorflow as tfimport numpy as npimport matplotlib.pyplot as plt# 添加神经层def add_layer(inputs, in_size, out_size, activation_function=None):    Weights = tf.Variable(tf.random_normal([in_size, out_size]))    biases = tf.Variable(tf.zeros([1, out_size]) + 0.1)    Wx_plus_b = tf.matmul(inputs, Weights) + biases # 没有被激活的值    if activation_function is None: # 说明是线性方程        outputs = Wx_plus_b    else:        outputs = activation_function(Wx_plus_b)    return outputsx_data = np.linspace(-1,1,300)[:,np.newaxis] # 后面的中括号用于加维度noise = np.random.normal(0,0.05,x_data.shape)y_data = np.square(x_data) - 0.5 + noisexs = tf.placeholder(tf.float32, [None, 1]) # None表示无论给多少个sample都可以ys = tf.placeholder(tf.float32, [None, 1])# 1 -> 10 -> 1l1 = add_layer(xs, 1, 10, activation_function=tf.nn.relu)prediction = add_layer(l1, 10, 1, activation_function=None)loss = tf.reduce_mean(   # 求平均数          tf.reduce_sum( # 对10个例子的差进行求和            tf.square(ys - prediction), reduction_indices=[1]))train_step = tf.train.GradientDescentOptimizer(0.1).minimize(loss)init = tf.global_variables_initializer()with tf.Session() as sess:    sess.run(init)    fig = plt.figure()    # 规定坐标轴上下界    v = [-2, 2, -2, 2]    plt.axis(v)    ax = fig.add_subplot(1,1,1)    ax.scatter(x_data,y_data)    plt.ion() # python 3.5+ 防止show()阻塞屏幕    plt.show()    for i in range(1000):        sess.run(train_step,feed_dict={xs:x_data, ys:y_data})        if i% 5==0:            # print(sess.run(loss,feed_dict={xs:x_data, ys:y_data}))            try:                ax.lines.remove(lines[0])            except Exception:                pass            prediction_value = sess.run(prediction,feed_dict={xs:x_data})            lines = ax.plot(x_data,prediction_value,'r-',lw=5)            plt.pause(0.1)

(8) 过拟合以及tensorboard

// tensorboard 使用方法// 1. cd到logs文件夹的父目录// 2. 使用以下命令tensorboard --logdir='logs'// 3. 用浏览器打开指定的网页地址Starting TensorBoard b'47' at http://0.0.0.0:6006// 也就是http://0.0.0.0:6006// 注意：Windows中tensorboard经常gg, 推荐使用Mac

import tensorflow as tffrom sklearn.datasets import load_digitsfrom sklearn.cross_validation import train_test_splitfrom sklearn.preprocessing import LabelBinarizer# load datadigits = load_digits()X = digits.datay = digits.targety = LabelBinarizer().fit_transform(y)X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=.3)# 添加神经层def add_layer(inputs, in_size, out_size, layer_name, activation_function=None):    Weights = tf.Variable(tf.random_normal([in_size, out_size]))    biases = tf.Variable(tf.zeros([1, out_size]) + 0.1)    Wx_plus_b = tf.matmul(inputs, Weights) + biases # 没有被激活的值    # Dropout 一部分的神经元，减小过拟合    Wx_plus_b = tf.nn.dropout(Wx_plus_b, keep_prob)    if activation_function is None: # 说明是线性方程        outputs = Wx_plus_b    else:        outputs = activation_function(Wx_plus_b)    tf.summary.histogram(layer_name+'/outputs', outputs) # ver 1.0 以后必须这么写    return outputskeep_prob = tf.placeholder(tf.float32) # 每次保留多少百分百的值xs = tf.placeholder(tf.float32, [None, 64]) # None表示无论给多少个sample都可以ys = tf.placeholder(tf.float32, [None, 10])# add output layerl1 = add_layer(xs, 64, 50, 'l1', activation_function=tf.nn.tanh)prediction = add_layer(l1, 50, 10, 'l2', activation_function=tf.nn.softmax)# the loss between prediction and real datacross_entropy = tf.reduce_mean(-tf.reduce_sum(ys*tf.log(prediction),                                              reduction_indices=[1])) # losstf.summary.scalar('loss', cross_entropy)train_step = tf.train.GradientDescentOptimizer(0.6).minimize(cross_entropy)with tf.Session() as sess:    merged = tf.summary.merge_all()    # summary writer goes in here    train_writer = tf.summary.FileWriter("logs/train",sess.graph)    test_writer  = tf.summary.FileWriter("logs/test",sess.graph)    sess.run(tf.global_variables_initializer())    for i in range(1000):        sess.run(train_step,feed_dict={xs:X_train, ys:y_train, keep_prob: 0.5})        if i% 50==0:            # record loss            train_result = sess.run(merged, feed_dict={xs:X_train, ys:y_train, keep_prob: 1}) #记录的时候不用drop任何东西            test_result  = sess.run(merged, feed_dict={xs:X_test, ys:y_test, keep_prob: 1})            train_writer.add_summary(train_result,i)            test_writer.add_summary(test_result,i)

(9) 卷积神经网络

参考以下文章
http://blog.csdn.net/wang_junjie/article/details/51317348

import tensorflow as tffrom tensorflow.examples.tutorials.mnist import  input_data# number 1 to 10 datamnist = input_data.read_data_sets('MNIST_data',one_hot=True)def compute_accuracy(v_xs, v_ys):    global prediction    y_pre = sess.run(prediction, feed_dict={xs:v_xs, keep_prob:1})    correct_prediction = tf.equal(tf.argmax(y_pre,1), tf.argmax(v_ys,1))    accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))    result = sess.run(accuracy, feed_dict={xs:v_xs, ys:v_ys, keep_prob:1})    return  resultdef weight_variable(shape):    initial = tf.truncated_normal(shape, stddev=0.1) # 产生随机变量    return tf.Variable(initial)def bias_variable(shape):    initial = tf.constant(0.1, shape=shape)    return tf.Variable(initial)def conv2d(x, W):    return tf.nn.conv2d(x, W, strides=[1,1,1,1], padding='SAME')    # x:图片的所有信息， W:权值， strides:卷积步长    # strides规定为一个[1, x_movement, y_movement, 1],    # 第一个和第四个值必须为1，第二个值为X方向步长，第三个值为Y方向的步长    # padding : (1)SAME  : 抽取包括边缘值以外的值，以0填充，卷积后图片大小和原图片一致    #           (2)VALID : 只抽取边缘内的值，卷积后图片大小比原图片小def max_pool_2x2(x): # Pooling技术保留更多的图片信息    return tf.nn.max_pool(x, ksize=[1,2,2,1], strides=[1,2,2,1], padding='SAME')    # ksize第一个和第四个值必须为1# define placeholder for inputs to nerworkxs = tf.placeholder(tf.float32,[None,784])ys = tf.placeholder(tf.float32,[None,10])keep_prob = tf.placeholder(tf.float32)x_image = tf.reshape(xs,[-1,28,28,1]) # -1:表示不管这个维度, 1:图片高度## conv1 layer ##W_conv1 = weight_variable([5,5,1,32]) # patch 5x5 卷积器， in_size 1 图片的输入高度, output_size 32b_conv1 = bias_variable([32])h_conv1 = tf.nn.relu(conv2d(x_image,W_conv1) + b_conv1) # relu:非线性化, output_size 28x28x32 采用SAME所以长宽不变 而卷积输出高度为32h_pool1 = max_pool_2x2(h_conv1) # output_size 14x14x32## conv2 layer ##W_conv2 = weight_variable([5,5,32,64]) # patch 5x5 卷积器， in_size 32 图片的输入高度, output_size 64b_conv2 = bias_variable([64])h_conv2 = tf.nn.relu(conv2d(h_pool1,W_conv2) + b_conv2) # relu:非线性化, output_size 14x14x64 采用SAME所以长宽不变 而卷积输出高度为32h_pool2 = max_pool_2x2(h_conv2) # output_size 7x7x64## func1 layer ##W_fc1 = weight_variable([7*7*64,1024])b_fc1 = bias_variable([1024])h_pool2_flat = tf.reshape(h_pool2,[-1,7*7*64]) # 数据一维化 [n_samples,7,7,64] -> [n_samples,7*7*64]h_fc1 = tf.nn.relu(tf.matmul(h_pool2_flat, W_fc1)+b_fc1)h_fc1_drop = tf.nn.dropout(h_fc1, keep_prob)## func2 layer ##W_fc2 = weight_variable([1024,10])b_fc2 = bias_variable([10])prediction = tf.nn.softmax(tf.matmul(h_fc1_drop, W_fc2)+b_fc2) # softmax classification算概率# the error between prediction and real datacross_entropy = tf.reduce_mean(-tf.reduce_sum(ys*tf.log(prediction), reduction_indices=[1]))train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)with tf.Session() as sess:    sess.run(tf.global_variables_initializer())    for i in range(1000):        batch_xs, batch_ys = mnist.train.next_batch(100)        sess.run(train_step, feed_dict={xs:batch_xs, ys:batch_ys, keep_prob:0.5})        if i % 50 == 0:            print(compute_accuracy(mnist.test.images, mnist.test.labels))

(10) 保存读取Weights和Bias

import tensorflow as tfimport numpy as np## Save to fileW = tf.Variable([[1,2,3],[3,4,5]], dtype=tf.float32, name='weights')b = tf.Variable([[1,2,3]], dtype=tf.float32, name='bias')saver = tf.train.Saver()with tf.Session() as sess:    sess.run(tf.global_variables_initializer())    # 一定要先建立一个my_net文件夹    save_path = saver.save(sess, "my_net/save_net.ckpt")    print("Save to path:", save_path)## Restore variables# restore时一定要定义dtype 和 shapeW = tf.Variable(np.arange(6).reshape((2,3)), dtype=tf.float32, name='weights')b = tf.Variable(np.arange(3).reshape((1,3)), dtype=tf.float32, name='bias')# not need init stepsaver = tf.train.Saver()with tf.Session() as sess:    saver.restore(sess,"my_net/save_net.ckpt")    print("weights:",sess.run(W))    print("bias:",sess.run(b))

(11) RNN

import tensorflow as tffrom tensorflow.examples.tutorials.mnist import input_data# number 1 to 10 datamnist = input_data.read_data_sets('MNIST_data',one_hot=True)#hyper parameterslr = 0.001training_iters = 100000batch_size = 128display_step = 10n_inputs = 28 # MNIST data input (img shape:28x28)n_steps  = 28 # time stepsn_hidden_units = 128 # neurons in hidden layern_classes = 10 # MNIST classes (0-9) 分为10个类# tf Graph inputx = tf.placeholder(tf.float32,[None, n_steps, n_inputs])y = tf.placeholder(tf.float32,[None, n_classes])# Define weightsweights = {    #(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,]))}def RNN(X, weights, biases):    # hidden layer for input to cell    # X(128batch, 28 steps, 28 inputs)    # ==> (128*28, 28 inputs)    X = tf.reshape(X,[-1,n_inputs])    # ==>(128batch * 28steps, 128 hidden)    X_in = tf.matmul(X, weights['in']) + biases['in']    # ==>(128batch , 28steps, 128 hidden)    X_in= tf.reshape(X_in,[-1,n_steps,n_hidden_units])    # cell    lstm_cell = tf.contrib.rnn.BasicLSTMCell(n_hidden_units, forget_bias=1.0, state_is_tuple=True)    _init_state = lstm_cell.zero_state(batch_size, dtype=tf.float32)    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    results = tf.matmul(states[1], weights['out']) + biases['out']    return resultspred = 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)correct_pred = tf.equal(tf.argmax(pred, 1), tf.argmax(y,1))accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))with tf.Session() as sess:    sess.run(tf.global_variables_initializer())    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 % 100 == 0:            print(sess.run(accuracy,feed_dict={                x: batch_xs,                y: batch_ys            }))