keras概述

准备

!pip3 install tensorflow==2.0.0a0
!pip3 install pyyaml
%matplotlib inline

Looking in indexes: https://pypi.tuna.tsinghua.edu.cn/simple
Requirement already satisfied: tensorflow==2.0.0a0 in /usr/local/lib/python3.7/site-packages (2.0.0a0)
Requirement already satisfied: keras-preprocessing>=1.0.5 in /usr/local/lib/python3.7/site-packages (from tensorflow==2.0.0a0) (1.0.9)
Requirement already satisfied: grpcio>=1.8.6 in /usr/local/lib/python3.7/site-packages (from tensorflow==2.0.0a0) (1.19.0)
Requirement already satisfied: tb-nightly<1.14.0a20190302,>=1.14.0a20190301 in /usr/local/lib/python3.7/site-packages (from tensorflow==2.0.0a0) (1.14.0a20190301)
Requirement already satisfied: gast>=0.2.0 in ./Library/Python/3.7/lib/python/site-packages (from tensorflow==2.0.0a0) (0.2.2)
Requirement already satisfied: astor>=0.6.0 in ./Library/Python/3.7/lib/python/site-packages (from tensorflow==2.0.0a0) (0.7.1)
Requirement already satisfied: termcolor>=1.1.0 in ./Library/Python/3.7/lib/python/site-packages (from tensorflow==2.0.0a0) (1.1.0)
Requirement already satisfied: wheel>=0.26 in /usr/local/lib/python3.7/site-packages (from tensorflow==2.0.0a0) (0.33.1)
Requirement already satisfied: google-pasta>=0.1.2 in /usr/local/lib/python3.7/site-packages (from tensorflow==2.0.0a0) (0.1.4)
Requirement already satisfied: tf-estimator-nightly<1.14.0.dev2019030116,>=1.14.0.dev2019030115 in /usr/local/lib/python3.7/site-packages (from tensorflow==2.0.0a0) (1.14.0.dev2019030115)
Requirement already satisfied: six>=1.10.0 in ./Library/Python/3.7/lib/python/site-packages (from tensorflow==2.0.0a0) (1.12.0)
Requirement already satisfied: protobuf>=3.6.1 in /usr/local/lib/python3.7/site-packages (from tensorflow==2.0.0a0) (3.7.0)
Requirement already satisfied: keras-applications>=1.0.6 in /usr/local/lib/python3.7/site-packages (from tensorflow==2.0.0a0) (1.0.7)
Requirement already satisfied: numpy<2.0,>=1.14.5 in /usr/local/lib/python3.7/site-packages (from tensorflow==2.0.0a0) (1.16.2)
Requirement already satisfied: absl-py>=0.7.0 in ./Library/Python/3.7/lib/python/site-packages (from tensorflow==2.0.0a0) (0.7.0)
Requirement already satisfied: werkzeug>=0.11.15 in ./Library/Python/3.7/lib/python/site-packages (from tb-nightly<1.14.0a20190302,>=1.14.0a20190301->tensorflow==2.0.0a0) (0.14.1)
Requirement already satisfied: markdown>=2.6.8 in ./Library/Python/3.7/lib/python/site-packages (from tb-nightly<1.14.0a20190302,>=1.14.0a20190301->tensorflow==2.0.0a0) (3.0.1)
Requirement already satisfied: setuptools in /usr/local/lib/python3.7/site-packages (from protobuf>=3.6.1->tensorflow==2.0.0a0) (40.8.0)
Requirement already satisfied: h5py in ./Library/Python/3.7/lib/python/site-packages (from keras-applications>=1.0.6->tensorflow==2.0.0a0) (2.9.0)
Looking in indexes: https://pypi.tuna.tsinghua.edu.cn/simple
Requirement already satisfied: pyyaml in /usr/local/lib/python3.7/site-packages (5.1)

导入tf.keras

tf.keras(除非特殊说明，本文在非代码中使用tf.keras表示tensorflow的keras，而直接使用keras则表示keras标准)是Tensorflow基于keras标准实现的模块，该模块提供了高层的API用于快速构建和训练模型，并且对Tensorflow中的eager执行器、tf.data和Estimators提供了支持。可以说tf.keras在不损失性能的前提下简化了Tensorflow的使用。
使用tf.keras，导入下面的模块即可。

from tensorflow import keras

tf.keras可以执行任意与Keras相符合的代码，但是需要注意下面两点：

1. 最新的Tensorflow中的tf.keras版本可能和PyPI中keras的版本不一样
2. 当使用tf.keras保存模型的时候，默认的格式是tf的checkpoint格式，如果指定save_format="h5"才会保存HDF5格式

创建简单模型

顺序模型

在Keras中，你可以通过收集不同的层创建模型，通常一个模型就是基于层构建的计算图。最为常见的模型就是通过堆叠不同的层来构建，此时就可以使用tf.keras.Sequential。
这里创建一个简单的全连接网络(多层感知器)

model = keras.Sequential()
# 添加一个64个神经元的全连接层
model.add(keras.layers.Dense(64, activation='relu'))
# 再次添加
model.add(keras.layers.Dense(64, activation='relu'))
# 添加一个10个神经元的softmax
model.add(keras.layers.Dense(10, activation='softmax'))

可以发现，使用tf.kerasAPI创建一个网络真的十分简单

层

tf.keras.layers中实现了常见的层：全连接Dense、卷积Conv2D、池化MaxPool2D等等，其他文章我会详细介绍。
这里简单介绍几个在tf.keras.layers中常用到的参数：

1. activation指定了激活函数对象，该参数可以是一个可调用函数或者一个字符串(最后还是转换为可调用函数)(字符串和函数的区别是，使用字符串无法指定函数的具体参数，只能使用默认的)，默认是没有激活函数
2. kernel_initializer和bias_initializer，这两个分别指定了权重和偏置的初始化函数对象，同样可以是一个可调用函数或者一个字符串，默认情况下是Glorot uniform
3. kernel_regularizer和bias_regularizer，提供了权重和偏置的正则化函数对象，默认是没有

下面示例代码演示和如何使用上面的参数：

# sigmoid激活
keras.layers.Dense(10, activation='sigmoid')
# 或者
keras.layers.Dense(10, activation=keras.activations.sigmoid)
# 对权重使用0.01的L1正则化
keras.layers.Dense(10, kernel_regularizer=keras.regularizers.l1(0.01))
# 对偏置使用0.01的L2正则化
keras.layers.Dense(10,bias_regularizer=keras.regularizers.l2(0.01))
# 对权重使用orthogonal进行初始化
keras.layers.Dense(10,kernel_initializer='orthogonal')
# 对偏置使用常量2.0进行初始化
keras.layers.Dense(10,bias_initializer=keras.initializers.Constant(2.0))

<tensorflow.python.keras.layers.core.Dense at 0x1090e4e10>

训练和评估

训练前准备

当模型构建完成后，可以通过调用compile方法，指定网络的训练配置

model = keras.Sequential([
    # 添加一个神经元个数为64的全连接，并指定输入为尺寸
    keras.layers.Dense(64, activation='relu', input_shape=(32, )), 
    # 再添加一层
    keras.layers.Dense(64, activation='relu'), 
    # 添加softmax
    keras.layers.Dense(10, activation='softmax')])

model.compile(optimizer=keras