Predictive Models

供个人学习记录，来源于：
https://github.com/machinelearningmindset/TensorFlow-Course#why-use-tensorflow

import os 
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2' 
import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt
import urllib    #操作URL
import tempfile    #临时文件和目录的处理
import pandas as pd   #解决数据分析任务
from tensorflow.examples.tutorials.mnist import input_data

max_num_checkpoint = 10
num_classes = 2
batch_size = 512
num_epochs = 10

initial_learning_rate = 0.001  #学习率
learning_rate_decay_factor = 0.95   #衰减率
num_epochs_per_decay = 1  #使用次数

is_training = False
fine_tuning = False
online_test = True
allow_soft_placement = True
log_device_placement = False

mnist = input_data.read_data_sets("MNIST_data/", reshape=True, one_hot=False)  #读取MNIST数据集

data={}

data['train/image'] = mnist.train.images
data['train/label'] = mnist.train.labels
data['test/image'] = mnist.test.images
data['test/label'] = mnist.test.labels

def extract_samples_Fn(data):  #选取0类和1类
  index_list = []
  for sample_index in range(data.shape[0]):
    label = data[sample_index]
    if label == 1 or label == 0:
      index_list.append(sample_index)
  return index_list

index_list_train = extract_samples_Fn(data['train/label'])
index_list_test = extract_samples_Fn(data['test/label'])

data['train/image'] = mnist.train.images[index_list_train]
data['train/label'] = mnist.train.labels[index_list_train]
data['test/image'] = mnist.test.images[index_list_test]
data['test/label'] = mnist.test.labels[index_list_test]

dimensionality_train = data['train/image'].shape

num_train_samples = dimensionality_train[0]   #训练数据量
num_features = dimensionality_train[1]    #训练图片大小

graph = tf.Graph()   #实例化一个类

with graph.as_default():    #创建一张默认图
  global_step = tf.Variable(0, name="global_step", trainable=False)   #用于衰减的全局步骤
  decay_steps = int(num_train_samples/batch_size*num_epochs_per_decay)  #衰减速度，每隔**衰减一次
  learning_rate = tf.train.exponential_decay(initial_learning_rate,global_step,decay_steps,learning_rate_decay_factor,staircase=True,name='exponential_decay_learning_rate')  #指数梯度下降法
  image_place = tf.placeholder(tf.float32, shape=([None, num_features]), name='image')
  label_place = tf.placeholder(tf.int32, shape=([None,]), name='gt')
  label_one_hot = tf.one_hot(label_place, depth=num_classes, axis=-1)   #label进行one_hot编码
  dropout_param = tf.placeholder(tf.float32)   #dropout
  
  logits = tf.contrib.layers.fully_connected(inputs=image_place, num_outputs = num_classes, scope='fc')  #定义全连接层

  with tf.name_scope('loss'):
    loss_tensor = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=logits, labels=label_one_hot))  #求交叉熵

  prediction_correct = tf.equal(tf.argmax(logits, 1), tf.argmax(label_one_hot, 1))    #行寻找最大值比较
  accuracy = tf.reduce_mean(tf.cast(prediction_correct, tf.float32))    #转换数据类型求均值
  optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate)    #梯度优化

  with tf.name_scope('train_op'):
    gradients_and_variables = optimizer.compute_gradients(loss_tensor)   #计算损失函数对变量的梯度
    train_op = optimizer.apply_gradients(gradients_and_variables, global_step=global_step)   #变量梯度更新

  session_conf = tf.ConfigProto(allow_soft_placement=allow_soft_placement,log_device_placement=log_device_placement)   #session参数配置
  sess = tf.Session(graph=graph, config=session_conf)

  with sess.as_default():    #创建一个默认会话
    saver = tf.train.Saver()   #参数保存
    sess.run(tf.global_variables_initializer())
    checkpoint_prefix = 'model'
    if fine_tuning:
      saver.restore(sess, os.path.join(checkpoint_path, checkpoint_prefix))   #提取训练好的参数
      print("Model restored for fine-tuning...")

    test_accuracy = 0
    for epoch in range(num_epochs):
      total_batch_training = int(data['train/image'].shape[0] / batch_size)
      for batch_num in range(total_batch_training):
        start_idx = batch_num * batch_size
        end_idx = (batch_num + 1) * batch_size
        train_batch_data, train_batch_label = data['train/image'][start_idx:end_idx], data['train/label'][start_idx:end_idx]
        batch_loss, _, training_step = sess.run([loss_tensor, train_op, global_step],feed_dict={image_place:train_batch_data,label_place:train_batch_label,dropout_param:0.5})
      print("Epoch " + str(epoch + 1) + ", Training Loss= " + "{:.5f}".format(batch_loss))

    test_accuracy = 100 * sess.run(accuracy, feed_dict={image_place: data['test/image'],label_place: data['test/label'],dropout_param: 1.})
    
    print("Final Test Accuracy is %% %.2f" % test_accuracy)