@Author：Runsen

多层感知机（MLP）有着非常悠久的历史，多层感知机(MLP)是深度神经网络(DNN)的基础算法

MLP基础知识

• 目的：创建用于简单回归/分类任务的常规神经网络（即多层感知器）和Keras

MLP结构

• 每个MLP模型由一个输入层、几个隐藏层和一个输出层组成
• 每层神经元的数目不受限制

回归任务的MLP

• 当目标（y）连续时
• 对于损失函数和评估指标，通常使用均方误差（MSE）

from tensorflow.keras.datasets import boston_housing
(X_train, y_train), (X_test, y_test) = boston_housing.load_data()

数据集描述

• 波士顿住房数据集共有506个数据实例（404个培训和102个测试）
• 13个属性（特征）预测“某一地点房屋的中值”
• 文件编号：https://keras.io/datasets/

1.创建模型

• Keras模型对象可以用Sequential类创建
• 一开始，模型本身是空的。它是通过添加附加层和编译来完成的
• 文档：https://keras.io/models/sequential/

from tensorflow.keras.models import Sequential

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
• 文档：https://keras.io/layers/core/

from tensorflow.keras.layers import Activation, Dense
# Keras model with two hidden layer with 10 neurons each 
model.add(Dense(10, input_shape = (13,)))    # Input layer => input_shape should be explicitly designated
model.add(Activation('sigmoid'))
model.add(Dense(10))                         # Hidden layer => only output dimension should be designated
model.add(Activation('sigmoid'))
model.add(Dense(10))                         # Hidden layer => only output dimension should be designated
model.add(Activation('sigmoid'))
model.add(Dense(1))                          # Output layer => output dimension = 1 since it is regression problem
# This is equivalent to the above code block
model.add(Dense(10, input_shape = (13,), activation = 'sigmoid'))
model.add(Dense(10, activation = 'sigmoid'))
model.add(Dense(10, activation = 'sigmoid'))
model.add(Dense(1))

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/

from tensorflow.keras import optimizers

sgd = optimizers.SGD(lr = 0.01)    # stochastic gradient descent optimizer

model.compile(optimizer = sgd, loss = 'mean_squared_error', metrics = ['mse'])    # for regression problems, mean squared error (MSE) is often employed

模型摘要

model.summary()

odel: "sequential"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense (Dense)                (None, 10)                140       
_________________________________________________________________
activation (Activation)      (None, 10)                0         
_________________________________________________________________
dense_1 (Dense)              (None, 10)                110       
_________________________________________________________________
activation_1 (Activation)    (None, 10)                0         
_________________________________________________________________
dense_2 (Dense)              (None, 10)                110       
_________________________________________________________________
activation_2 (Activation)    (None, 10)                0         
_________________________________________________________________
dense_3 (Dense)              (None, 1)                 11        
_________________________________________________________________
dense_4 (Dense)              (None, 10)                20        
_________________________________________________________________
dense_5 (Dense)              (None, 10)                110       
_________________________________________________________________
dense_6 (Dense)              (None, 10)                110       
_________________________________________________________________
dense_7 (Dense)              (None, 1)                 11        
=================================================================
Total params: 622
Trainable params: 622
Non-trainable params: 0
_________________________________________________________________

2.培训

• 使用提供的训练数据训练模型

model.fit(X_train, y_train, batch_size = 50, epochs = 100, verbose = 1)

3.评估

• Keras模型可以用evaluate（）函数计算
• 评估结果包含在列表中
• 文档：https://keras.io/metrics/

results = model.evaluate(X_test, y_test)

print(model.metrics_names)     # list of metric names the model is employing
print(results)                 # actual figure of metrics computed

print('loss: ', results[0])
print('mse: ', results[1])