fashion_mnist数据集分类模型及详细注解笔记

可以用ipython编程，方便查看代码块的输出结果，代码如下：

import tensorflow as tf
import matplotlib as mpl
import matplotlib.pyplot as plt
%matplotlib inline
import numpy as np
import sklearn
import pandas as pd
import os
import sys
import time 

from tensorflow import keras

print(tf.__version__)
print(sys.version_info)
for module in mpl,np,pd,sklearn,tf,keras:
    print(module.__name__,module.__version__)#打印版本信息

输出：
2.0.0
sys.version_info(major=3, minor=7, micro=3, releaselevel=‘final’, serial=0)
matplotlib 3.1.1
numpy 1.17.3
pandas 0.25.3
sklearn 0.21.3
tensorflow 2.0.0
tensorflow_core.keras 2.2.4-tf
数据集处理：

fashion_mnist = keras.datasets.fashion_mnist#下载数据集
(x_train_all,y_train_all),(x_test,y_test) = fashion_mnist.load_data()
x_valid,x_train = x_train_all[:5000],x_train_all[5000:]
y_valid,y_train = y_train_all[:5000],y_train_all[5000:]
#共60000张图片，前5000张分给验证集，后55000张分给训练集

print(x_valid.shape,y_valid.shape)
print(x_train.shape,y_train.shape)
print(x_test.shape,y_test.shape)#打印数据集维度

输出：

(5000, 28, 28) (5000,)
(55000, 28, 28) (55000,)
(10000, 28, 28) (10000,)

print(np.max(x_train),np.min(x_train))#打印归一化之前训练集的最大值和最小值

输出：

255 0

#归一化提高准确率：
#x=(x-u)/std

from sklearn.preprocessing import StandardScaler

scaler = StandardScaler()
#fit_transform把训练集转化为归一化后的数据，把x_train三维[None,28,28]变成二维[None，784]，然后再reshape变回来，fit可以记录训练集均值方差
x_train_scaled = scaler.fit_transform(
                 x_train.astype(np.float32).reshape(-1, 1)).reshape(-1, 28, 28)
x_valid_scaled = scaler.transform(
                 x_valid.astype(np.float32).reshape(-1, 1)).reshape(-1, 28, 28)
x_test_scaled = scaler.transform(
                 x_test.astype(np.float32).reshape(-1, 1)).reshape(-1, 28, 28)

print(np.max(x_train_scaled),np.min(x_train_scaled))#打印归一化之后训练集的最大值和最小值

输出：
2.0231433 -0.8105136

定义一个显示单个图片的函数：

def show_single_image(i):
    plt.imshow(i,cmap="binary")
    plt.show()
    
show_single_image(x_train[0])#调用函数，显示训练集中的第一张图片

定义一个显示多个图片的函数：

def show_imgs(n_rows,n_cols,x_data,y_data,class_names):
    assert len(x_data) == len(y_data)
    assert n_rows * n_cols < len(x_data)#行和列的乘积不能大于数据总数
    plt.figure(figsize = (n_cols * 1,n_rows * 2))#定义大图
    for row in range(n_rows):
        for col in range(n_cols):
            index = n_cols * row + col
            plt.subplot(n_rows, n_cols, index+1)#显示子图
            plt.imshow(x_data[index],cmap = "binary",interpolation = 'nearest')#显示图片，最近邻值插值
            plt.axis('off')#去掉坐标系
            plt.title(class_names[y_data[index]])
    plt.show()
    
class_names = ['1','2','3','4','5','6','7','8','9','10']#类别名
show_imgs(3, 5,x_valid,y_valid,class_names)#调用函数显示大图，3行5列数据

输出：

模型部分：

model = keras.models.Sequential()
model.add(keras.layers.Flatten(input_shape = [28,28]))#将28*28的二维向量展平为一维向量784
model.add(keras.layers.Dense(400,activation = 'relu'))#全连接层
model.add(keras.layers.Dense(200,activation = 'relu'))
model.add(keras.layers.Dense(10,activation = 'softmax'))#使用softmax函数进行分类，分为10类
#为了提高准确率，模型中的参数可以自己改

#上面模型还可以这样子写：
'''
model=keras.models.Sequential([
    keras.layers.Flatten(input_shape = [28,28]),
    keras.layers.Dense(400,activation = 'relu'),
    keras.layers.Dense(200,activation = 'relu'),
    keras.layers.Dense(10,activation = 'softmax')'''
])
#relu: y = max(0,x)
#softmax ； 将向量变成概率分布，x=[x1,x2,x3]
#           y=[e^x1/sum,e^x2/sum,e^x3/sum],sum=e^x1+e^x2+e^x3

#sparse:把y--index变成one_hot编码
model.compile(loss = 'sparse_categorical_crossentropy',
              optimizer = "adam",
             metrics = ["accuracy"])

model.summary()#查看模型概况，参数
#y=wx+b=[400,784]*[784]+[400]
#784*400+400=314000

输出：

开始训练数据：

history=model.fit(x_train_scaled,y_train,epochs=10,
                 validation_data=(x_valid_scaled,y_valid))

训练结果如下，准确率大概可以达到0.87

查看history函数类型：

type(history)

输出：
tensorflow.python.keras.callbacks.History

history.history#调用history函数，查看保存的历史数据，保存了损失值，准确率，验证集损失值，验证集准确率

输出：
{‘loss’: [2.3015895209962673,
0.4815958445809104,
0.44026698377782647,
0.43106005086465315,
0.4143984176288952,
0.40346119379130274,
0.3827689492702484,
0.37051838736967607,
0.3626356901017102,
0.3551486981045116],
‘accuracy’: [0.77312726,
0.83116364,
0.8431818,
0.8449818,
0.8536909,
0.8559091,
0.86463636,
0.86856365,
0.87307274,
0.87438184],
‘val_loss’: [0.5049341471433639,
0.47250081301927566,
0.4290150542378426,
0.3783755409121513,
0.533234388539195,
0.5099339093625546,
0.40061412162780763,
0.3903807807683945,
0.4614589884161949,
0.3808077231764793],
‘val_accuracy’: [0.8176,
0.8392,
0.8532,
0.8666,
0.8228,
0.8268,
0.8666,
0.8736,
0.8422,
0.8734]}

画出结果曲线：

def plot_learning_curves(history):
    pd.DataFrame(history.history).plot(figsize = (8,5))#dataframe取数据，图的大小为8*5
    plt.grid(True)#设置网格
    plt.gca().set_ylim(0,1)#设置y轴坐标轴范围
    plt.show()#显示数据
    
plot_learning_curves(history)

输出：

测试集评估：

model.evaluate(x_test_scaled,y_test)

结果：