【深度学习】实验07 使用TensorFlow完成逻辑回归

使用tensorflow完成逻辑回归

tensorflow是一种开源的机器学习框架，由google brain团队于2015年开发。它被广泛应用于图像和语音识别、自然语言处理、推荐系统等领域。

tensorflow的核心是用于计算的数据流图。在数据流图中，节点表示数学操作，边表示张量（多维数组）。将操作和数据组合在一起的数据流图可以使 tensorflow 对复杂的数学模型进行优化，同时支持分布式计算。

tensorflow提供了python，c++，java，go等多种编程语言的接口，让开发者可以更便捷地使用tensorflow构建和训练深度学习模型。此外，tensorflow还具有丰富的工具和库，包括tensorboard可视化工具、tensorflow serving用于生产环境的模型服务、keras高层封装api等。

tensorflow已经发展出了许多优秀的模型，如卷积神经网络、循环神经网络、生成对抗网络等。这些模型已经在许多领域取得了优秀的成果，如图像识别、语音识别、自然语言处理等。

除了开源的tensorflow，google还推出了基于tensorflow的云端机器学习平台google cloud ml，为用户提供了更便捷的训练和部署机器学习模型的服务。

解决分类问题里最普遍的baseline model就是逻辑回归，简单同时可解释性好，使得它大受欢迎，我们来用tensorflow完成这个模型的搭建。

1. 环境设定

import os
os.environ['tf_cpp_min_log_level']='2'

import warnings
warnings.filterwarnings("ignore")

import numpy as np
import tensorflow as tf
from tensorflow.examples.tutorials.mnist import input_data
import time

2. 数据读取

#使用tensorflow自带的工具加载mnist手写数字集合
mnist = input_data.read_data_sets('./data/mnist', one_hot=true) 

extracting ./data/mnist/train-images-idx3-ubyte.gz
extracting ./data/mnist/train-labels-idx1-ubyte.gz
extracting ./data/mnist/t10k-images-idx3-ubyte.gz
extracting ./data/mnist/t10k-labels-idx1-ubyte.gz

#查看一下数据维度
mnist.train.images.shape

(55000, 784)

#查看target维度
mnist.train.labels.shape

(55000, 10)

3. 准备好placeholder

batch_size = 128
x = tf.placeholder(tf.float32, [batch_size, 784], name='x_placeholder') 
y = tf.placeholder(tf.int32, [batch_size, 10], name='y_placeholder')

4. 准备好参数/权重

w = tf.variable(tf.random_normal(shape=[784, 10], stddev=0.01), name='weights')
b = tf.variable(tf.zeros([1, 10]), name="bias")

logits = tf.matmul(x, w) + b 

5. 计算多分类softmax的loss function

# 求交叉熵损失
entropy = tf.nn.softmax_cross_entropy_with_logits(logits=logits, labels=y, name='loss')
# 求平均
loss = tf.reduce_mean(entropy)

6. 准备好optimizer

这里的最优化用的是随机梯度下降，我们可以选择adamoptimizer这样的优化器

learning_rate = 0.01
optimizer = tf.train.adamoptimizer(learning_rate).minimize(loss)

7. 在session里执行graph里定义的运算

#迭代总轮次
n_epochs = 30

with tf.session() as sess:
    # 在tensorboard里可以看到图的结构
    writer = tf.summary.filewriter('../graphs/logistic_reg', sess.graph)

    start_time = time.time()
    sess.run(tf.global_variables_initializer())	
    n_batches = int(mnist.train.num_examples/batch_size)
    for i in range(n_epochs): # 迭代这么多轮
        total_loss = 0
        for _ in range(n_batches):
            x_batch, y_batch = mnist.train.next_batch(batch_size)
            _, loss_batch = sess.run([optimizer, loss], feed_dict={x: x_batch, y:y_batch}) 
            total_loss += loss_batch
        print('average loss epoch {0}: {1}'.format(i, total_loss/n_batches))
    print('total time: {0} seconds'.format(time.time() - start_time))
    print('optimization finished!')

# 测试模型
    preds = tf.nn.softmax(logits)
    correct_preds = tf.equal(tf.argmax(preds, 1), tf.argmax(y, 1))
    accuracy = tf.reduce_sum(tf.cast(correct_preds, tf.float32))
    
    n_batches = int(mnist.test.num_examples/batch_size)
    total_correct_preds = 0
    
    for i in range(n_batches):
        x_batch, y_batch = mnist.test.next_batch(batch_size)
        accuracy_batch = sess.run([accuracy], feed_dict={x: x_batch, y:y_batch}) 
        total_correct_preds += accuracy_batch[0]
        
    print('accuracy {0}'.format(total_correct_preds/mnist.test.num_examples))

    writer.close()

   average loss epoch 0: 0.36748782022571785    
   average loss epoch 1: 0.2978815356126198    
   average loss epoch 2: 0.27840628396797845    
   average loss epoch 3: 0.2783186247437706    
   average loss epoch 4: 0.2783641471138923    
   average loss epoch 5: 0.2750668214473413           
   average loss epoch 6: 0.2687560408126502    
   average loss epoch 7: 0.2713795114126239    
   average loss epoch 8: 0.2657588795522154    
   average loss epoch 9: 0.26322007090686916    
   average loss epoch 10: 0.26289192279735646    
   average loss epoch 11: 0.26248606019989873       
   average loss epoch 12: 0.2604622903056356    
   average loss epoch 13: 0.26015280702939403    
   average loss epoch 14: 0.2581879366319496    
   average loss epoch 15: 0.2590309207117085    
   average loss epoch 16: 0.2630510463581219    
   average loss epoch 17: 0.25501730025578767    
   average loss epoch 18: 0.2547102673000945    
   average loss epoch 19: 0.258298404375851    
   average loss epoch 20: 0.2549241428330784    
   average loss epoch 21: 0.2546788509283866    
   average loss epoch 22: 0.259556887067837    
   average loss epoch 23: 0.25428259843365575    
   average loss epoch 24: 0.25442713139565676    
   average loss epoch 25: 0.2553852511383159    
   average loss epoch 26: 0.2503043229415978    
   average loss epoch 27: 0.25468004046828596    
   average loss epoch 28: 0.2552785321479633    
   average loss epoch 29: 0.2506257003663859    
   total time: 28.603315353393555 seconds    
   optimization finished!    
   accuracy 0.9187

附：系列文章