卷积神经网络_1.基本cnn网络实现（使用relu激活函数）

import matplotlib as mpl  # Matplotlib 是 Python 的绘图库。 它可与 NumPy 一起使用

import matplotlib.pyplot as plt  # Python数据可视化matplotlib.pyplot

# %matplotlib inline #在使用jupyter notebook 或者 jupyter qtconsole的时候，经常会用到%matplotlib inline。其作用就是在你调用plot()进行画图或者直接输入Figure的实例对象的时候，会自动的显示并把figure嵌入到console中。

import numpy as np  # 数值计算扩展。这种工具可用来存储和处理大型矩阵

import sklearn  # 机器学习中常用的第三方模块，对常用的机器学习方法进行了封装，包括回归(Regression)、降维(Dimensionality Reduction)、分类(Classfication)、聚类(Clustering)等方法。

import pandas as pd  # 是python的一个数据分析包
import os  # 系统编程的操作模块，可以处理文件和目录
import sys  # sys模块包含了与Python解释器和它的环境有关的函数
import time
import tensorflow as tf

from tensorflow import keras
##################################################################################################
# 选择GPU
os.environ["CUDA_VISIBLE_DEVICES"] = "0"

##################################################################################################

print(tf.__version__)
print(sys.version_info)
for module in mpl, np, pd, sklearn, tf, keras:
    print(module.__name__, module.__version__)

##################################################################################################


fashion_mnist = keras.datasets.fashion_mnist  # 从keras中导入数据集datasets

(x_train_all, y_train_all), (x_test, y_test) = fashion_mnist.load_data(  )# load_data导入训练集和测试集,x是图片，y是label

x_valid, x_train = x_train_all[:5000], x_train_all[5000:  ]  # 导入数据集中前x_valid5000个和x_train 为5000之后的
y_valid, y_train = y_train_all[:5000], y_train_all[5000:  ]  # 导入数据集中前y_valid5000个和y_train 为5000之后的

print(x_valid.shape, y_valid.shape  )  # 前5000个训练集
print(x_train.shape, y_train.shape  )  # 训练集
print(x_test.shape, y_test.shape  )  # 测试集

##################################################################################################
# x = (x - u) / std u是均值，std是方差
# 符合均值是0，方差是1的正态分布

from sklearn.preprocessing import StandardScaler

scaler = StandardScaler()

x_train_scaled = scaler.fit_transform(
    x_train.astype(np.float32).reshape(-1, 1)).reshape(-1, 28, 28,1)
x_valid_scaled = scaler.transform(
    x_valid.astype(np.float32).reshape(-1, 1)).reshape(-1, 28, 28,1)
x_test_scaled = scaler.transform(
    x_test.astype(np.float32).reshape(-1, 1)).reshape(-1, 28, 28,1)

# 归一化之前的accurary0.1多，归一化之后的accurary0.9多

##################################################################################################

model = keras.models.Sequential()
#添加两个卷积层一个pooling层
model.add(keras.layers.Conv2D(filters=32,kernel_size =3,padding = 'same',activation= 'relu',input_shape=(28,28,1)))
model.add(keras.layers.Conv2D(filters=32,kernel_size =3,padding = 'same',activation= 'relu'))
model.add(keras.layers.MaxPool2D(pool_size = 2))

#添加两个卷积层一个pooling层
model.add(keras.layers.Conv2D(filters=64,kernel_size =3,padding = 'same',activation= 'relu'))
model.add(keras.layers.Conv2D(filters=64,kernel_size =3,padding = 'same',activation= 'relu'))
model.add(keras.layers.MaxPool2D(pool_size = 2))

#添加两个卷积层一个pooling层
model.add(keras.layers.Conv2D(filters=128,kernel_size =3,padding = 'same',activation= 'relu'))
model.add(keras.layers.Conv2D(filters=128,kernel_size =3,padding = 'same',activation= 'relu'))
model.add(keras.layers.MaxPool2D(pool_size = 2))

#添加全连接层
#先flatten
model.add(keras.layers.Flatten())
model.add(keras.layers.Dense(128,activation='relu'))
model.add(keras.layers.Dense(10, activation="softmax"))

#filters 有多个少个卷积核（输出的通道)
#kernal_size 卷积核的 大小
#padding 边缘补零，same表示输入经过卷积之后输出和输入的大小一样



##################################################################################################

# 配置训练模型
model.compile(loss="sparse_categorical_crossentropy",  # 损失函数
              optimizer = "sgd",  # 优化器名
              metrics = ["accuracy"])
# reason for sparse: y->index. y->one_hot->[]
# 原因：y是长度等于样本数目的向量，对于每个样本来说都是一个值，sparse使用y变成一个向量
##################################################################################################
print(model.layers)
"""output:
    [, , , , , , , , , , , ]
    
"""

print(model.summary())
"""output:
    Model: "sequential"
    _________________________________________________________________
    Layer (type)                 Output Shape              Param #   
    =================================================================
    conv2d (Conv2D)              (None, 28, 28, 32)        320       
    _________________________________________________________________
    conv2d_1 (Conv2D)            (None, 28, 28, 32)        9248      
    _________________________________________________________________
    max_pooling2d (MaxPooling2D) (None, 14, 14, 32)        0         
    _________________________________________________________________
    conv2d_2 (Conv2D)            (None, 14, 14, 64)        18496     
    _________________________________________________________________
    conv2d_3 (Conv2D)            (None, 14, 14, 64)        36928     
    _________________________________________________________________
    max_pooling2d_1 (MaxPooling2 (None, 7, 7, 64)          0         
    _________________________________________________________________
    conv2d_4 (Conv2D)            (None, 7, 7, 128)         73856     
    _________________________________________________________________
    conv2d_5 (Conv2D)            (None, 7, 7, 128)         147584    
    _________________________________________________________________
    max_pooling2d_2 (MaxPooling2 (None, 3, 3, 128)         0         
    _________________________________________________________________
    flatten (Flatten)            (None, 1152)              0         
    _________________________________________________________________
    dense (Dense)                (None, 128)               147584    
    _________________________________________________________________
    dense_1 (Dense)              (None, 10)                1290      
    =================================================================
    Total params: 435,306
    Trainable params: 435,306
    Non-trainable params: 0
    _________________________________________________________________
    None
    
    Process finished with exit code 0

"""

##################################################################################################
# 回调函数 Tensorboard, earlystopping, ModelCheckpoint
# logdir = './keras实战/callbacks'#这样运行会报错，可能是在windows下路径导致的问题
logdir = "cnn-selu-callbacks"
if not os.path.exists(logdir):
    os.mkdir(logdir)
output_model_file = os.path.join(logdir ,"fashion_mnist_model.h5")

callbacks = [
    keras.callbacks.TensorBoard(logdir)  ,  # 保存在logdir中
    keras.callbacks.ModelCheckpoint(output_model_file,  # 模型名称
                                    save_best_only = True),  # 保存一个最好的模型
    keras.callbacks.EarlyStopping(patience=5, min_delta=1e-3)  ,  # 如果连续5次迭代，loss值达到最小误差1e-3，便停止训练
]


##################################################################################################
# 开始训练
history = model.fit(x_train_scaled, y_train, epochs=10,
                    validation_data=(x_valid_scaled, y_valid),
                    callbacks = callbacks)


##################################################################################################
# 绘制结果图
def plot_learning_curves(history):
    pd.DataFrame(history.history).plot(figsize=(8 ,5))
    plt.grid(True)
    plt.gca().set_ylim(0 ,3)
    plt.show()


plot_learning_curves(history)
# selu也可以缓解梯度消失
# 标准化的好处可以使椭圆分布的数据（梯度下降时难以拟合出最佳路线）呈现圆的分布

##################################################################################################

# 用训练好的model进行评估
model.evaluate(x_test_scaled ,y_test)
"""output:
    1s 109us/sample - loss: 0.1774 - accuracy: 0.8934
"""