当前位置:网站首页>Neural Network Study Notes 4 - Autoencoder (including sparse, stacked) (updated)
Neural Network Study Notes 4 - Autoencoder (including sparse, stacked) (updated)
2022-07-30 10:40:00 【Oreos are delicious】
目录
配套讲解视频
It is recommended to read the blog post in conjunction with the video
10Minutes to learn automatic encoder from principle to programming implementation_哔哩哔哩_bilibili
10Minutes to learn automatic encoder from principle to programming implementation
1.Programs and Datasets
链接:https://pan.baidu.com/s/1aSNq94BJuKsiKO5gNGF29Q
提取码:6666
--来自百度网盘超级会员V5的分享
2.自动编码器
2.1自编码器原理
Learning efficient encodings in a set of data by means of unsupervised learning


The target of the autoencoder:Re-extract features,降低维度,Minimize refactoring errors
目标函数:Make the refactoring error as 0

Simple two-layer autoencoder example

2.2代码实现
1.数据导入
D=xlsread('C:\Users\86188\Desktop\B站ppt\ae\RaisinDataset.xlsx');
data=D(:,1:7)';
label=D(:,8)';2.数据集处理
k=rand(1,900);
[m,n]=sort(k);
input_train=data(:,n(1:750));
input_test=data(:,n(751:900));
output_train=label(:,n(1:750));
output_test=label(:,n(751:900));
x = input_train;
t=ind2vec(output_train);
t=full(t)3.设置网络结构
ae1=trainAutoencoder(x,5);
features=encode(ae1,x);
softmax=trainSoftmaxLayer(features,t);
nets=stack(ae1,softmax);4.显示结果,测试网络
view(nets)
output_test
y=nets(input_test) 
5.计算准确率
for i=1:150
output_fore(i)=find(y(:,i)==max(y(:,i)));
end
right1=0;
for i=1:150
if output_fore(i) == output_test(i)
right1=right1+1;
end
end
right=right1/1503.堆叠式自编码器
对于很多数据来说,仅使用两层神经网络的自编码器还不足以获取一种好的数据表示,为了获取更好的数据表示,我们可以使用更深层的神经网络,深层神经网络作为自编码器提取的数据表示一般会更加抽象,It can better capture the semantic information of the data.
在实践中经常使用逐层堆叠的方式来训练一个深层的自编码器.称为堆叠自编码器(StackedAuto-EncoderSAE)Stacked autoencoders can generally be trained layer by layer(Layer-WiseTraining)来学习网络参数
4.稀疏自编码器
4.1稀疏编码
Advantages of sparse coding
(1).计算量
稀疏性带来的最大好处就是可以极大地降低计算量.
(2)可解释性
因为稀疏编码只有少数的非零元素,相当于将一个输入样本表示为少数几个相关的特征.这样我们可以更好地描述其特征,并易于理解.
(3)特征选择
稀疏性带来的另外一个好处是可以实现特征的自动选择,只选择和输入样本相关的最少特征,This allows for a better representation of the input samples,降低噪声并减轻过拟合.
4.2.稀疏自编码器
By feeding the hidden layer units in the autoencoderz加上稀疏性限制,自编码器可以学习到数据中一些有用的结构.
目标函数

WRepresents a parameter in the autoencoder
和稀疏编码一样,The advantage of sparse autoencoders is high interpretability,并同时进行了隐式的特征选择.
结构图

边栏推荐
- JVM内存布局、类加载机制及垃圾回收机制详解
- MySQL | Subqueries
- Re17:读论文 Challenges for Information Extraction from Dialogue in Criminal Law
- 4. yolov5-6.0 ERROR: AttributeError: 'Upsample' object has no attribute 'recompute_scale_factor' solution
- 元宇宙改变人类工作模式的四种方式
- flowable workflow all business concepts
- Do you really understand the 5 basic data structures of Redis?
- Study Notes 10--Main Methods of Local Trajectory Generation
- A near-perfect Unity full-platform hot update solution
- js对数组操作移动进行封装
猜你喜欢

神经网络学习笔记4——自动编码器(含稀疏,堆叠)(更新中)

再有人问你分布式事务,把这篇扔给他

ospf2 two-point two-way republish (question 2)

死锁的理解

(C语言)文件操作

spark udf accepts and handles null values.

从数据流中快速查找中位数

SST-Calib: A lidar-visual extrinsic parameter calibration method combining semantics and VO for spatiotemporal synchronization calibration (ITSC 2022)

In the robot industry professionals, Mr Robot industry current situation?

Quick Start Tutorial for flyway
随机推荐
Soft test system architects introductory tutorial | system operation and software maintenance
Detailed explanation of JVM memory layout, class loading mechanism and garbage collection mechanism
Mysterious APT Attack
Study Notes 10--Main Methods of Local Trajectory Generation
4. yolov5-6.0 ERROR: AttributeError: 'Upsample' object has no attribute 'recompute_scale_factor' solution
第2章 常用安全工具
MFCC to audio, the effect should not be too funny >V
Re18:读论文 GCI Everything Has a Cause: Leveraging Causal Inference in Legal Text Analysis
idea2021+Activiti【最完整笔记一(基础使用)】
(BUG record) No module named PIL
【HMS core】【Analytics Kit】【FAQ】如何解决华为分析付费分析中付款金额显示为0的问题?
ospf2 two-point two-way republish (question 2)
Re17: Read the paper Challenges for Information Extraction from Dialogue in Criminal Law
Meikle Studio - see the actual combat notes of Hongmeng device development 4 - kernel development
Security思想项目总结
你真的懂Redis的5种基本数据结构吗?
再有人问你分布式事务,把这篇扔给他
Practical Walkthrough | Calculate Daily Average Date or Time Interval in MySQL
Study Notes 11--Direct Construction of Local Trajectories
多线程--线程和线程池的用法