第二章：张量基本介绍+一个基本框架_现有的张量处理框架

from keras.datasets import mnist 
(train_images, train_labels), (test_images, test_labels) = mnist.load_data()
# 1.张量的三个属性
print(train_images.ndim)    # 轴的个数 3
print(train_images.shape)   # 形状   （60000,28,28）
print(train_images.dtype)   # 数据类型  unit8
 
# 看看其中第4张图的样子 
digit = train_images[4]
import matplotlib.pyplot as plt
plt.imshow(digit, cmap = plt.cm.binary)
plt.show()
 
# 2.张量操作
my_slice = train_images[10:100]            # 沿着第一个轴选取第10~100个图片（不包括100）
print(my_slice.shape)                      # (90,28,28)
my_slice = train_images[10:100,:,:]        # 等价操作
my_slice = train_images[10:100，0:28,0:28] # 等价操作
 
# 3.一般来说数据张量的第一个轴（0轴）都是样本轴
batch = train_images[:128]                 # 数据集的一个批量，大小为128
# 3.现实世界的数据张量
'''
1.向量数据：2D张量，形状为（samples,features)
2.时间序列数据或序列数据：3D张量，形状为(sample,timesteps,features)
3.图像：4D张量，形状为(sample,height,width,channels)or(sample,channels,height,width)
4.视频：5D张量，形状为(sample,frame,height,width,channels)or(samples,frames,channels,height,width)
'''
 
# 4.张量运算
# 4.1 numpy中元素运算
import numpy as np
x = np.array([1,2,3])
y = np.array([4,5,-4])
z = x+y                      # 逐个元素相加
print(z)                     # [5,7,-1]
z = np.maximum(z,0)          # 逐元素relu
print(z)                     # [5,7,0]
# 4.2 广播
x = np.random.random((64, 3, 32, 10))    
y = np.random.random((32, 10))    
z = np.maximum(x, y)                      # y的（32,10）重复了（64,3）
                                          # z的输出为64,3,32,10的张量
# 4.3张量点积
x.shape[1] == y.shape[0]
np.dot(x,y)                                          
# 4.4张量变形
train_images = train_images.reshape((60000,28*28))
x = np.zeros((300,20))
x = np.transpose(x)                       # 变形之转置
 
# 网络架构
from keras import models
from keras import layers
network = models.Sequential()
network.add(layers.Dense(512,activation='relu',input_shape = (28*28,)))        # dense代表全连接层=512我理解的是输出是512神经元在这一层
network.add(layers.Dense(10,activation = 'softmax'))                           # 输出10个神经元
#编译：损失函数，优化器，评价指标
network.compile(optimizer='rmsprop',loss='categorical_crossentropy',metrics=['accuracy'])
#数据预处理 取值范围在0，1而且是float 32
train_images = train_images.reshape((60000, 28 * 28)) 
train_images = train_images.astype('float32') / 255 
test_images = test_images.reshape((10000, 28 * 28)) 
test_images = test_images.astype('float32') / 255 
#准备标签
from keras.utils import to_categorical 
train_labels = to_categorical(train_labels)
test_labels = to_categorical(test_labels) 
#模型拟合（会显示之前定义的损失loss和metrics）
network.fit(train_images, train_labels, epochs=5, batch_size=128) 
#模型测试
test_loss, test_acc = network.evaluate(test_images, test_labels) 
print('test_acc:', test_acc)