TensorFlow框架介绍:
TensorFlow是谷歌基于DistBelief(第一代人工智能学习系统)研发的第二代人工智能学习系统。
AlphaGo就是基于TensorFlow研发的。
TensorFlow支持CNN、RNN 和 LSTM算法。
TensorFlow 从1.0升级为2.0后做了一些改变,具体可查看tensorflow的官方文档:
MNIST数据集介绍:
MNIST数据集是一个存放手写数字(0-9)的数据库。
该数据集中有70000张训练图像(60000张训练图像和10000张的测试图像)
展现形式:
MNIST数据集模型搭建:
pycharm新建项目minist_testdemo (conda环境)
新建mnist文件夹
mnist文件夹下新建MNIST_data文件夹和data文件夹
将下载的四个包直接放入MNIST_data中(不用解压)
导入训练数据:
mnist文件夹下新建input_data.py
from __future__import absolute_import
from __future__import division
from __future__import print_function
import gzip
import os
import tempfile
import numpy
from six.movesimport urllib
from six.movesimport xrange
import tensorflow.compat.v1as tf
tf.disable_v2_behavior()
#导入mnist数据集
from tensorflow.keras.datasets.mnistimport load_data
定义训练模型:
mnist文件夹下新建model.py,定义线性模型和卷积模型
import tensorflow.compat.v1as tf
tf.disable_v2_behavior()
#定义线性模型
#Y=W*x+b
def regression(x):
#784*10的图像
W = tf.Variable(tf.zeros([784, 10]), name="W")
#维度为10
b = tf.Variable(tf.zeros([10]), name="b")
#y=wx+b
y = tf.nn.softmax(tf.matmul(x, W) + b)
return y, [W, b]
#定义卷积模型
#多重卷积
def convolutional(x, keep_prob):
#定义卷积层2*2
def conv2d(x, W):
return tf.nn.conv2d(x, W, [1, 1, 1, 1], padding='SAME')
#定义2*2池化层
def max_pool_2x2(x):
return tf.nn.max_pool(x, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding='SAME')
#定义权重变量
def weight_variable(shape):
initial = tf.truncated_normal(shape, stddev=0.1)
return tf.Variable(initial)
#
def bias_variable(shape):
initial = tf.constant(0.1, shape=shape)
return tf.Variable(initial)
#定义卷积重,图像张量
#第一重
x_image = tf.reshape(x, [-1, 28, 28, 1])
W_conv1 = weight_variable([5, 5, 1, 32])
b_conv1 = bias_variable([32])
h_conv1 = tf.nn.relu(conv2d(x_image, W_conv1) + b_conv1)
h_pool1 = max_pool_2x2(h_conv1)
#第二重
W_conv2 = weight_variable([5, 5, 32, 64])
b_conv2 = bias_variable([64])
h_conv2 = tf.nn.relu(conv2d(h_pool1, W_conv2) + b_conv2)
h_pool2 = max_pool_2x2(h_conv2)
#full connection
#全连接层
W_fc1 = weight_variable([7 *7 *64, 1024])
b_fc1 = bias_variable([1024])
h_pool2_flat = tf.reshape(h_pool2, [-1, 7*7*64])
h_fc1 = tf.nn.relu(tf.matmul(h_pool2_flat, W_fc1) + b_fc1)
#防止过拟合
h_fc1_drop = tf.nn.dropout(h_fc1, keep_prob)
W_fc2 = weight_variable([1024, 10])
b_fc2 = bias_variable([10])
y = tf.nn.softmax(tf.matmul(h_fc1_drop, W_fc2) + b_fc2)
return y, [W_conv1, b_conv1, W_conv2, b_conv2, W_fc1, b_fc1, W_fc2, b_fc2]
使用线性模型进行训练:
mnist文件夹下新建regression.py
import os
import input_data
import model
import tensorflow.compat.v1as tf
tf.disable_v2_behavior()
data = input_data.load_data('MNIST_data')
# create model
# 引入参数
with tf.variable_scope("regression"):
x = tf.placeholder(tf.float32, [None, 784])
y, variables = model.regression(x)
# train
y_ = tf.placeholder("float", [None, 10])
# 交叉熵
cross_entropy = -tf.reduce_sum(y_ * tf.log(y))
# 训练步骤
train_step = tf.train.GradientDescentOptimizer(0.01).minimize(cross_entropy)
# 预测
correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(y_, 1))
# 计算准确率
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
# 保存参数
saver = tf.train.Saver(variables)
# 开始训练
with tf.Session()as sess:
# 将所有参数进行全局初始化放进来
sess.run(tf.global_variables_initializer())
# 训练1000次
for _in range(1000):
batch_xs, batch_ys = data.train.next_batch(100)
# feed_dict:往里放参数
sess.run(train_step, feed_dict={x: batch_xs, y_: batch_ys})
# 打印准确度和测试数据集的images,labels
print((sess.run(accuracy, feed_dict={x: data.test.images, y_: data.test.labels})))
# 保存模型
path = saver.save(
sess, os.path.join(os.path.dirname(__file__), 'data', 'regression.ckpt'),
write_meta_graph=False, write_state=False)
#打印模型的保存路径
print("Saved:", path)
使用卷积模型进行训练
mnist文件夹下新建convolutional.py
import os
import model
import tensorflow.compat.v1as tf
import input_data
tf.disable_v2_behavior()
#导入数据集
data = input_data.load_data('MNIST_data')
#定义model
with tf.variable_scope("convolutional"):
x = tf.placeholder(tf.float32, [None, 784], name='x')
keep_prob = tf.placeholder(tf.float32)
y, variables = model.convolutional(x, keep_prob)
#train
#张量大小为[none,10]
y_ = tf.placeholder(tf.float32, [None, 10], name='y')
#交叉熵
cross_entropy = -tf.reduce_sum(y_ * tf.log(y))
#训练步长,使用随机梯度下降
train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
#预测
correct_prediction = tf.equal(tf.argmax(y, 1), tf.argmax(y_, 1))
#准确率
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
#保存参数
saver = tf.train.Saver(variables)
#进行训练
with tf.Session()as sess:
#回归参数
merged_summary_op = tf.summary.merge_all()
#
summay_writer = tf.summary.FileWriter('/tmp/mnist_log/1', sess.graph)
#把图加进来
summay_writer.add_graph(sess.graph)
sess.run(tf.global_variables_initializer())
#循环20000次
#训练数据
for iin range(20000):
batch = data.train.next_batch(50)
#每隔100次打印一次准确率
if i %100 ==0:
train_accuracy = accuracy.eval(feed_dict={x: batch[0], y_: batch[1], keep_prob:1.0})
print("step %d, training accuracy %g" % (i, train_accuracy))
sess.run(train_step, feed_dict={x: batch[0], y_: batch[1], keep_prob:0.5})
#测试数据
print(sess.run(accuracy, feed_dict={x: data.test.images, y_: data.test.labels, keep_prob:1.0}))
#保存所有参数,路径
path = saver.save(
sess, os.path.join(os.path.dirname(__file__), 'data', 'convalutional.ckpt'),
write_meta_graph=False, write_state=False)
print("Saved:", path)