TensorFlow 2.0 学习笔记--第三章使用八股搭建神经网络

第三章使用八股搭建神经网络

本章目标：

神经网络搭建八股
iris代码复现
MNIST数据集
训练MNIST数据集
Fashion数据集

3.1 搭建网络八股 Sequential

用 Tensorflow API: tf.keras 搭建网络八股

六步法
- import
- train, test # 指定训练集特征和标签
- model = tf.keras.models.Sequential # 逐层描述网络
- model.compile # 使用哪种优化器、损失函数、评测指标等
- model.fit # 训练过程，告知训练集输入特征标签，batch，迭代次数
- model.summray # 打印出网络结构和参数统计
model = tf.keras.models.Sequential([网络结构]) # 描述各层网络

网络结构举例：
- 拉直层：tf.keras.layers.Flatten()
- 全连接层：tf.keras.layers.Dense(神经元个数, activation="激活函数", kernel_regularizer=哪种正则化)
  
  activation('可选字符串'): relu, softmax, sigmoid, tanh
  kernel_regularizer可选: tf.keras.regularizers.l1(), tf.kers.regularizers.l2()
- 卷积层：tf.keras.layers.Conv2D(filters=卷积核个数, kernel_size=卷积核尺寸, strides=卷积步长, padding="valid" or "same")
- LSTM层：tf.keras.layers.LSTM()
model.compile(optimizer=优化器, loss=损失函数, metrics=['准确率'])

Optimizer可选：(使用左边字符串或右边函数形式，自定义学习率等)
- 'sgd' or tf.keras.optimizers.SGD(lr=学习率, momentum=动量参数)
- 'adagrad' or tf.keras.optimizers.Adagrad(lr=学习率)
- 'adadelta' or tf.keras.optimizers.Adadelta(lr=学习率)
- 'adam' or tf.keras.optimizers.Adam(lr=学习率, beta_1=0.9, beta_2=0.999)
loss可选：
- 'mse' or tf,keras.losses.MeanSquaredError()
- 'sparse_categorical_crossentropy' or tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False) # 在询问是否是原始输出，经过概率分布为False，没有经过概率分布直接输出为True
Metrics可选：
- 'accuracy': y_ 和 y 都是数值，如 y_=[1] y=[1]
- 'categorical_accuracy': y_ 和 y 都是独热码（概率分布），如 y_=[0, 1, 0] y=[0.256, 0.695, 0.048]
- 'sparse_categorical_accuracy': y_ 是数值，y 是独热码（概率分布），如 y_=[1] y=[0.256, 0.695, 0.048]
model.fit()

model.fit(训练集的输入特征, 训练集的标签, batch_size= , epocher= , validation_data=(测试集的输入特征，测试集的标签), validation_split=从训练集划分多少比例给测试集, validation_freq=多少次epoch测试一次)

⭕ validation_data, validation_split 二选其一

model.summary() # 打印出网络的结构和参数统计

Model: "sequential"
_________________________________________________________________
Layer (type)                 Output Shape              Param #
=================================================================
dense (Dense)                (None, 3)                 15
=================================================================
Total params: 15
Trainable params: 15
Non-trainable params: 0
_________________________________________________________________

上述为一个4输入层，3输出层的网络。 (4 * 3 + 3 = 15)

p8_iris_sequential

import tensorflow as tf
from sklearn import datasets
import numpy as np

x_train = datasets.load_iris().data
y_train = datasets.load_iris().target

np.random.seed(116)
np.random.shuffle(x_train)
np.random.seed(116)
np.random.shuffle(y_train)
tf.random.set_seed(116)

model = tf.keras.models.Sequential([
    tf.keras.layers.Dense(3, activation='softmax', kernel_regularizer=tf.keras.regularizers.l2())
])

model.compile(optimizer=tf.keras.optimizers.SGD(lr=0.1),
              loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False),
              metrics=['sparse_categorical_accuracy'])

model.fit(x_train, y_train, batch_size=32, epochs=500, validation_split=0.2, validation_freq=20)

model.summary()

# 运行结果
# Epoch 498/500
# 4/4 [==============================] - 0s 1ms/step - loss: 0.3509 - sparse_categorical_accuracy: 0.9658
# Epoch 499/500
# 4/4 [==============================] - 0s 1ms/step - loss: 0.3691 - sparse_categorical_accuracy: 0.9306
# Epoch 500/500
# 4/4 [==============================] - 0s 14ms/step - loss: 0.3634 - sparse_categorical_accuracy: 0.9304 - val_loss: 0.3516 - val_sparse_categorical_accuracy: 0.8667
# Model: "sequential"
# _________________________________________________________________
# Layer (type)                 Output Shape              Param #   
# =================================================================
# dense (Dense)                (None, 3)                 15        
# =================================================================
# Total params: 15
# Trainable params: 15
# Non-trainable params: 0
# _________________________________________________________________

3.2 搭建网络八股 class

用Tensorflow API: tf.keras 搭建网络八股

class MyModel(Model):  # Model表示继承了TensorFlow的Model类
    def __init__(self):
        super(MyModel, self).__init__()
        定义网络结构块
    def call(self, x):  # x 为输入数据
        调用网络结构块，实现前向传播
        return y
model = MyModel()

__init__() 定义所欲要网络结构块
call() 写出前向传播

iris_class 代码

# 与iris_sequential不同之处
from tensorflow.keras.layers import Dense
from tensorflow.keras import Model

class IrisModel(Model):
    def __init__(self):
        super(IrisModel, self).__init__()
        self.d1 = Dense(3, activation='softmax', kernel_regularizer=tf.keras.regularizers.l2())

    def call(self, x):
        y = self.d1(x)
        return y

model = IrisModel()

# 运行结果
# 同iris_Sequential

3.3 MNIST数据集：

提供6万张 28 * 28 像素点的 0~9 手写数字图片和标签，用于训练。
提供1万张 28 * 28 像素点的 0~9 手写数字图片和标签，用于测试。

导入MNIST数据集：

mnist = tf.keras.datasets.mnist
(x_train, y_train), (x_test, y_test) = mnist.load_data()

作为输入特征，输入神经网络时，将数据拉伸为一维数组：

tf.keras.layers.Flatten()
# [0 0 0 48 238 252 252 ... ... ... 253 186 12 0 0 0 0 0 0]

可视化各个数据以及完整代码

import tensorflow as tf
from matplotlib import pyplot as plt

mnist = tf.keras.datasets.mnist
(x_train, y_train), (x_test, y_test) = mnist.load_data()

# 可视化训练集输入特征的第一个元素
plt.imshow(x_train[0], cmap='gray')  # 绘制灰度图
plt.show()

# 打印出训练集输入特征的第一个元素
print("x_train[0]:\n", x_train[0])
# 打印出训练集标签的第一个元素
print("y_train[0]:\n", y_train[0])

# 打印出整个训练集输入特征形状
print("x_train.shape:\n", x_train.shape)
# 打印出整个训练集标签的形状
print("y_train.shape:\n", y_train.shape)
# 打印出整个测试集输入特征的形状
print("x_test.shape:\n", x_test.shape)
# 打印出整个测试集标签的形状
print("y_test.shape:\n", y_test.shape)

# 运行结果
# x_train[0]:
#  '一个二维数组'
# y_train[0]:
#  5
# x_train.shape:
#  (60000, 28, 28)
# y_train.shape:
#  (60000,)
# x_test.shape:
#  (10000, 28, 28)
# y_test.shape:
#  (10000,)

💬 Codes 完整的Mnist手写数据集代码

import tensorflow as tf

mnist = tf.keras.datasets.mnist
(x_train, y_train), (x_test, y_test) = mnist.load_data()
x_train, x_test = x_train / 255.0, x_test / 255.0  # 数值变小后更利于神经网络的吸收

model = tf.keras.models.Sequential([
    tf.keras.layers.Flatten(),  # 将数据集拉伸为一维
    tf.keras.layers.Dense(128, activation='relu'),  # 第一个Dense是隐藏层
    tf.keras.layers.Dense(10, activation='softmax')  # 第二个Dense是输出层
])

model.compile(
    optimizer='adam',
    loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False),
    metrics=['sparse_categorical_accuracy']
)

model.fit(x_train, y_train, batch_size=32, epochs=5, validation_data=(x_test, y_test), validation_freq=1)
model.summary()


# 运行结果
# Epoch 1/5
# 2021-07-13 13:32:15.085136: I tensorflow/stream_executor/platform/default/dso_loader.cc:49] Successfully opened dynamic library libcublas.so.10
# 1875/1875 [==============================] - 4s 2ms/step - loss: 0.4395 - sparse_categorical_accuracy: 0.8750 - val_loss: 0.1424 - val_sparse_categorical_accuracy: 0.9567
# Epoch 2/5
# 1875/1875 [==============================] - 4s 2ms/step - loss: 0.1232 - sparse_categorical_accuracy: 0.9626 - val_loss: 0.1036 - val_sparse_categorical_accuracy: 0.9679
# Epoch 3/5
# 1875/1875 [==============================] - 4s 2ms/step - loss: 0.0828 - sparse_categorical_accuracy: 0.9749 - val_loss: 0.0934 - val_sparse_categorical_accuracy: 0.9701
# Epoch 4/5
# 1875/1875 [==============================] - 3s 2ms/step - loss: 0.0582 - sparse_categorical_accuracy: 0.9827 - val_loss: 0.0789 - val_sparse_categorical_accuracy: 0.9742
# Epoch 5/5
# 1875/1875 [==============================] - 4s 2ms/step - loss: 0.0432 - sparse_categorical_accuracy: 0.9870 - val_loss: 0.0768 - val_sparse_categorical_accuracy: 0.9767
# Model: "sequential"
# _________________________________________________________________
# Layer (type)                 Output Shape              Param #   
# =================================================================
# flatten (Flatten)            (None, 784)               0         
# _________________________________________________________________
# dense (Dense)                (None, 128)               100480    
# _________________________________________________________________
# dense_1 (Dense)              (None, 10)                1290      
# =================================================================
# Total params: 101,770
# Trainable params: 101,770
# Non-trainable params: 0
# _________________________________________________________________

# class
from tensorflow.keras.layers import Dense, Flatten
from tensorflow.keras import Model

class MnistModel(Model):
    def __init__(self):
        super(MnistModel, self).__init__()
        self.flatten = Flatten()
        self.d1 = Dense(128, activation='relu')
        self.d2 = Dense(10, activation='softmax')

    def call(self, x):
        x = self.flatten(x)
        x = self.d1(x)
        y = self.d2(x)
        return y

3.4 FASHION数据集：

提供6万张 28 * 28 像素点的衣裤等图片和标签，用于训练。
提供1万张 28 * 28 像素点的衣裤等图片和标签，用于测试。

导入FASHION数据集：

fashion = tf.keras.datasets.fashion_mnist
(x_train, y_train), (x_test, y_test) = fashion.load_data()

💬 Codes

import tensorflow as tf

fashion = tf.keras.datasets.fashion_mnist
(x_train, y_train), (x_test, y_test) = fashion.load_data()
x_train, x_test = x_train / 255.0, x_test / 255.0

model = tf.keras.models.Sequential([
    tf.keras.layers.Flatten(),
    tf.keras.layers.Dense(128, activation='relu'),
    tf.keras.layers.Dense(10, activation='softmax')
])

model.compile(
    optimizer='adam',
    loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=False),
    metrics=['sparse_categorical_accuracy']
)

model.fit(x_train, y_train, batch_size=32, epochs=5, validation_data=(x_test, y_test), validation_freq=1)
model.summary()

# 运行结果
# 记得自己动手 try

posted @ 2021-07-13 13:38 AlsoRan 阅读(158) 评论(0) 收藏举报

刷新页面返回顶部

AlsoRan

TensorFlow 2.0 学习笔记--第三章 使用八股搭建神经网络

第三章 使用八股搭建神经网络