tensorflow 调用mnist.npz数据集 手写数字识别 逻辑回归方法

本方法及程序适用于tensorflow1.X

mnist.npz与mnist.gz区别

mnist.gz来源于tensorflow

包含四个文件：
train-images-idx3-ubyte.gz
train-labels-idx1-ubyte.gz
t10k-images-idx3-ubyte.gz
/t10k-labels-idx1-ubyte.gz

from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets('data/', one_hot=True)
train_images = mnist.train.images
train_labels = mnist.train.labels
test_images  = mnist.test.images
test_labels  = mnist.test.labels

图像size：
train_images：（55000，784）#784=28X28，图像像素点拉成一维
test_images：（10000，784）
train_labels：（55000，10）#10维中，第n个为1其余为0，表示数字n
test_labels：（10000，10）

mnist.npz来源于keras

包含一个文件：
mnist.npz

from tensorflow.keras.datasets import mnist
(train_images, train_labels), (test_images, test_labels) = mnist.load_data()

图像size：
train_images：（60000，28，28）
test_images：（10000，28，28）
train_labels：（60000）
test_labels：（10000）

由于tensorflow已经把keras作为一个核心API，且tensorflow原先的mnist.gz数据集在使用时会有warning，如想继续用tensorflow的方法做手写数字识别，且使用mnist.npz数据集，可用下述代码。
mnist.npz数据集不连接外网下不了，可自行搜索资源进行下载。另外下载之后为了避免程序运行时再去请求外网下载，可更改mnist.py文件（目录：anaconda安装位置+\Lib\site-packages\tensorflow_core\python\keras\datasets）进行配置

代码

将mnist.npz的数据size转化为mnist.gz的size（并归一化）

import tensorflow.compat.v1 as tf
from tensorflow.keras.datasets import mnist
from tensorflow import keras
import Method

#加载数据集并规范化
(train_images, train_labels), (test_images, test_labels) = mnist.load_data()

train_images = train_images.reshape(60000,784)/255
train_labels = keras.utils.to_categorical(train_labels,10)

test_images = test_images.reshape(10000,784)/255
test_labels = keras.utils.to_categorical(test_labels,10)

print("MNIST dataset loaded")
print(train_images.shape,test_images.shape) # out: (60000, 28, 28)
print(train_labels.shape,test_labels.shape) # out: (60000,)

Method类之后会提到，用于存放next_batch函数。

变量定义及变量初始化

#参数初始化
x = tf.placeholder("float",[None,784])
y = tf.placeholder("float",[None,10])
W = tf.Variable(tf.zeros([784,10]))
b = tf.Variable(tf.zeros([10]))

#逻辑回归模型
actv = tf.nn.softmax(tf.matmul(x,W)+b)
#损失函数
cost = tf.reduce_mean(-tf.reduce_sum(y*tf.log(actv),reduction_indices=1))
#优化
learning_rate = 0.01
optm = tf.train.GradientDescentOptimizer(learning_rate).minimize(cost)

#预测
pred = tf.equal(tf.argmax(actv,1),tf.argmax(y,1))
#准确率
accr = tf.reduce_mean(tf.cast(pred,"float"))
#初始化
init = tf.global_variables_initializer()

迭代回归

training_epochs = 50
batch_size = 100
display_step = 5
#session
sess = tf.Session()
sess.run(init)
#mini-batch-learning
for epoch in range(training_epochs):
    avg_cost = 0
    num_batch = int(train_images.shape[0]/batch_size)
    for i in range(num_batch):
        batch_xs, batch_ys = Method.next_batch(train_images,train_labels,batch_size)
        #print(batch_xs, batch_ys)
        #batch_xs,batch_ys = train_images,train_labels
        sess.run(optm,feed_dict = {x:batch_xs, y:batch_ys})
        feeds = {x:batch_xs, y:batch_ys}
        avg_cost += sess.run(cost, feed_dict=feeds)/num_batch
        #print(i)
        #print(sess.run(cost, feed_dict=feeds))
    #display
    if epoch % display_step == 0:
        feeds_train = {x:batch_xs, y:batch_ys}
        feeds_test = {x:test_images, y:test_labels}
        train_acc = sess.run(accr,feed_dict = feeds_train)
        test_acc = sess.run(accr,feed_dict = feeds_test)
        print("Epoch: %03d/%03d cost: %.9f train_acc: %.3f test_acc: %.3f"
              % (epoch, training_epochs, avg_cost, train_acc, test_acc))
print ("DONE")

next_batch实现

由于使用tensorflow之前的mnist.gz会有warning，但是keras没找到next_batch类似函数。可新建一个Method类，里面包含如下next_batch函数，引用自他人程序

# 随机取batch_size个训练样本
import numpy as np

def next_batch(train_data, train_target, batch_size):
    index = [ i for i in range(0,len(train_target)) ]
    np.random.shuffle(index);
    batch_data = [];
    batch_target = [];
    for i in range(0,batch_size):
        batch_data.append(train_data[index[i]]);
        batch_target.append(train_target[index[i]])
    return batch_data, batch_target

如上，可实现tensorflow完成mnist.npz数据集的逻辑回归算法。

——冬至