利用mnist数据集做深度学习小实例
# -*- coding: utf-8 -*-"""Created on Fri Apr 20 09:03:09 2018@author: Administrator"""import tensorflow as tf import numpy as npfrom tensorflow.examples.tutorials.mnist import input_dataimport os os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'#对一个张量进行全面的汇总(均值，标准差，最大最小值，直方图)# (用于 TensorBoard 可视化).def variable_summaries(var): with tf.name_scope('summaries'): mean = tf.reduce_mean(var) tf.summary.scalar('mean', mean) with tf.name_scope('stddev'): stddev = tf.sqrt(tf.reduce_mean(tf.square(var - mean))) tf.summary.scalar('stddev', stddev) tf.summary.scalar('max', tf.reduce_max(var)) tf.summary.scalar('min', tf.reduce_min(var)) tf.summary.histogram('histogram', var)print('~~~~~~~~~~开始设计计算图~~~~~~~~')# 告诉TensorFlow模型将会被构建在默认的Graph上.with tf.Graph().as_default(): #定义输入节点 with tf.name_scope('DataInput'): X = tf.placeholder(tf.float32, shape=[None, 784], name='X') Y_true = tf.placeholder(tf.float32, shape=[None, 10], name='Y_true') #将数据转换为图片 with tf.name_scope('input_reshape'): image_shaped_input = tf.reshape(X, [-1, 28, 28, 1]) tf.summary.image('input', image_shaped_input, 10) #定义前项输入过程 with tf.name_scope('Interface'): W = tf.Variable(tf.zeros([784,10]), name='Wights') variable_summaries(W) b = tf.Variable(tf.zeros([10]), name='Bias') variable_summaries(b) Y_pred = tf.add(tf.matmul(X, W), b) tf.summary.histogram('pred', Y_pred) #定义损失节点 with tf.name_scope('Loss'): #使用Softmax将前项预测结果转换为概率 with tf.name_scope('Softmax'): Y_pred_prob = tf.nn.softmax(logits=Y_pred) tf.summary.histogram('pred_prob', Y_pred_prob) #使用交叉熵损失 with tf.name_scope('cross_entropy'): TrainLoss = tf.reduce_mean(-tf.reduce_sum(Y_true*tf.log(Y_pred_prob),axis=1)) tf.summary.scalar('cross_entropy', TrainLoss) #定义训练节点 with tf.name_scope('Train'): # Optimizer: 创建一个梯度下降优化器 Optimizer = tf.train.GradientDescentOptimizer(learning_rate=0.01) # Train: 定义训练节点将梯度下降法应用于Loss TrainStep = Optimizer.minimize(TrainLoss) #Evaluate: 定义评估节点 with tf.name_scope('Evaluate'): correct_prediction = tf.equal(tf.argmax(Y_pred_prob, 1), tf.argmax(Y_true, 1)) accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32)) tf.summary.scalar('accuracy', accuracy) #整合所有节点，写入事件文件 merged = tf.summary.merge_all() # Initial:添加所有Variable类型的变量的初始化节点 InitOp = tf.global_variables_initializer() print('把计算图写入事件文件，在TensorBoard里面查看') train_writer = tf.summary.FileWriter(logdir='logs/mnist_softmax', graph=tf.get_default_graph()) #writer.close() #训练过程 mnist = input_data.read_data_sets('MNIST_data', one_hot=True)#获取数据 #启动会化 sess = tf.Session() sess.run(InitOp)#初始化所有变量W,b # 开始按批次训练，总共训练1000个批次，每个批次100个样本 for i in range(1000): batch_xs, batch_ys = mnist.train.next_batch(100) summary, _, train_loss= sess.run([merged, TrainStep, TrainLoss], feed_dict={X:batch_xs, Y_true: batch_ys}) #print("train step: ", i, ", train_loss: ", train_loss) train_writer.add_summary(summary, i) #对模型进行评估 accuracy_score = sess.run(accuracy,feed_dict={X: mnist.test.images, Y_true: mnist.test.labels}) #print("模型准确率：", accuracy_score) train_writer.close()
评论