【小白学习keras教程】二、基于CIFAR-10数据集训练简单的MLP分类模型

2022-08-18 09:33:47 浏览数 (1)

「@Author：Runsen」

分类任务的MLP

  • 当目标（「y」）是离散的（分类的）
  • 对于损失函数，使用交叉熵；对于评估指标，通常使用accuracy

数据集描述

  • CIFAR-10数据集包含10个类中的60000个图像—50000个用于培训，10000个用于测试
  • 有关更多信息，请参阅官方文档
代码语言：javascript复制
from tensorflow.keras.datasets import cifar10
from tensorflow.keras.utils import to_categorical
# load data and flatten X data to fit into MLP
(x_train, y_train), (x_test, y_test) = cifar10.load_data()
x_train, x_test = x_train.reshape(x_train.shape[0], -1), x_test.reshape(x_test.shape[0], -1)
y_train, y_test = to_categorical(y_train), to_categorical(y_test)
print(x_train.shape, x_test.shape, y_train.shape, y_test.shape)

1.创建模型

  • 与回归模型相同-使用Sequentia()
代码语言：javascript复制
model = Sequential()

1-1.添加层

  • Keras层可以「添加」到模型中
  • 添加层就像一个接一个地堆叠乐高积木
  • 应注意的是，由于这是一个分类问题，应添加sigmoid层（针对多类问题的softmax）
  • 文档：https://keras.io/layers/core/
代码语言：javascript复制
# Keras model with two hidden layer with 10 neurons each 
model.add(Dense(50, input_shape = (x_train.shape[-1],)))    # Input layer => input_shape should be explicitly designated
model.add(Activation('sigmoid'))
model.add(Dense(50))                         # Hidden layer => only output dimension should be designated
model.add(Activation('sigmoid'))
model.add(Dense(50))                         # Hidden layer => only output dimension should be designated
model.add(Activation('sigmoid'))
model.add(Dense(10))                          # Output layer => output dimension = 1 since it is regression problem
model.add(Activation('sigmoid'))
# This is equivalent to the above code block
model.add(Dense(50, input_shape = (x_train.shape[-1],), activation = 'sigmoid'))
model.add(Dense(50, activation = 'sigmoid'))
model.add(Dense(50, activation = 'sigmoid'))
model.add(Dense(10, activation = 'sigmoid'))

1-2.模型编译

  • Keras模型应在培训前“编译”
  • 应指定损失类型（函数）和优化器
  • 文档（优化器）：https://keras.io/optimizers/
  • 文档（损失）：https://keras.io/losses/
代码语言：javascript复制
from tensorflow.keras import optimizers
sgd = optimizers.SGD(lr = 0.01)    # stochastic gradient descent optimizer
model.compile(optimizer = sgd, loss = 'categorical_crossentropy', metrics = ['accuracy'])

模型摘要

代码语言：javascript复制
model.summary()
代码语言：javascript复制
Model: "sequential"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense (Dense)                (None, 50)                153650    
_________________________________________________________________
activation (Activation)      (None, 50)                0         
_________________________________________________________________
dense_1 (Dense)              (None, 50)                2550      
_________________________________________________________________
activation_1 (Activation)    (None, 50)                0         
_________________________________________________________________
dense_2 (Dense)              (None, 50)                2550      
_________________________________________________________________
activation_2 (Activation)    (None, 50)                0         
_________________________________________________________________
dense_3 (Dense)              (None, 10)                510       
_________________________________________________________________
activation_3 (Activation)    (None, 10)                0         
=================================================================
Total params: 159,260
Trainable params: 159,260
Non-trainable params: 0
_________________________________________________________________

2.训练

  • 使用提供的训练数据训练模型
代码语言：javascript复制
model.fit(x_train, y_train, batch_size = 128, epochs = 50, verbose = 1)

3.评估

  • Keras模型可以用evaluate（）函数计算
  • 文档：https://keras.io/metrics/
代码语言：javascript复制
results = model.evaluate(x_test, y_test)

在这里插入图片描述

代码语言：javascript复制
print(model.metrics_names)     # list of metric names the model is employing
print(results)                 # actual figure of metrics computed
https 网络安全 keras 打包

0 人点赞