机器学习——深度学习(Deep Learning)

以你之姓@ 2022-08-05 12:27 294阅读 0赞

Deep Learning是机器学习中一个非常接近AI的领域，其动机在于建立、模拟人脑进行分析学习的神经网络，最近研究了机器学习中一些深度学习的相关知识，本文给出一些很有用的资料和心得。

Key Words：有监督学习与无监督学习，分类、回归，密度估计、聚类，深度学习，Sparse DBN，

1. 有监督学习和无监督学习

给定一组数据(input，target)为Z=(X，Y)。

有监督学习：最常见的是regression & classification。

regression：Y是实数vector。回归问题，就是拟合(X，Y)的一条曲线，使得下式cost function L最小。

![1343977609_3572.png][]

classification：Y是一个finite number，可以看做类标号。分类问题需要首先给定有label的数据训练分类器，故属于有监督学习过程。分类问题中，cost function L(X,Y)是X属于类Y的概率的负对数。

![f458b0f65eb24e4d0f03630d0c18e607a61b0603.png][]，其中fi(X)=P(Y=i | X);

![ab2c31dc00245ab18699df4af7ecfab088ef692e.png][]

无监督学习：无监督学习的目的是学习一个function f，使它可以描述给定数据的位置分布P(Z)。 包括两种：density estimation & clustering.

density estimation就是密度估计，估计该数据在任意位置的分布密度

clustering就是聚类，将Z聚集几类（如K-Means），或者给出一个样本属于每一类的概率。由于不需要事先根据训练数据去train聚类器，故属于无监督学习。

PCA和很多deep learning算法都属于无监督学习。

2. [深度学习Deep Learning介绍][Deep Learning]

Depth 概念：depth: the length of the longest path from an input to an output.

Deep Architecture 的三个特点：深度不足会出现问题；人脑具有一个深度结构（每深入一层进行一次abstraction，由lower-layer的features描述而成的feature构成，就是[上篇中提到的feature hierarchy问题][feature hierarchy]，而且该hierarchy是一个[稀疏矩阵][Link 1]）；认知过程逐层进行，逐步抽象

[ 3篇文章][3]介绍Deep Belief Networks，作为DBN的breakthrough

3.Deep Learning Algorithm 的核心思想：

把learning hierarchy 看做一个network，则

①无监督学习用于每一层网络的pre-train；

②每次用无监督学习只训练一层，将其训练结果作为其higher一层的输入；

③用监督学习去调整所有层

这里不负责任地理解下，举个例子在Autoencoder中，无监督学习学的是feature，有监督学习用在fine-tuning. 比如每一个neural network 学出的hidden layer就是feature，作为下一次神经网络无监督学习的input……这样一次次就学出了一个deep的网络，每一层都是上一次学习的hidden layer。再用softmax classifier去fine-tuning这个deep network的系数。

![1343714901_3624.png][]

这三个点是Deep Learning Algorithm的精髓，我在[上一篇][feature hierarchy]文章中也有讲到，其中第三部分：Learning Features Hierachy & Sparse DBN就讲了如何运用Sparse DBN进行feature学习。

4. Deep Learning 经典阅读材料：

>  *  The monograph or review paper [Learning Deep Architectures for AI][] (Foundations & Trends in Machine Learning, 2009).
>  *  The ICML 2009 Workshop on Learning Feature Hierarchies [webpage][] has a [list of references][].
>  *  The LISA [public wiki][] has a [reading list][] and a [bibliography][].
>  *  Geoff Hinton has [readings][] from last year’s [NIPS tutorial][].
> 
> 阐述Deep learning主要思想的三篇文章：  
> 
> 
>  *  Hinton, G. E., Osindero, S. and Teh, Y., [A fast learning algorithm for deep belief nets][]Neural Computation 18:1527-1554, 2006  
>     
>  *  Yoshua Bengio, Pascal Lamblin, Dan Popovici and Hugo Larochelle, [Greedy Layer-Wise Training of Deep Networks][], in J. Platt et al. (Eds), Advances in Neural Information Processing Systems 19 (NIPS 2006), pp. 153-160, MIT Press, 2007**<比较了RBM和Auto-encoder>**  
>     
>  *  Marc’Aurelio Ranzato, Christopher Poultney, Sumit Chopra and Yann LeCun [Efficient Learning of Sparse Representations with an Energy-Based Model][], in J. Platt et al. (Eds), Advances in Neural Information Processing Systems (NIPS 2006), MIT Press, 2007**<将稀疏自编码用于回旋结构(convolutional architecture)>**  
>     
> 
> 06年后，大批deep learning文章涌现，感兴趣的可以看下大牛Yoshua Bengio的综述[Learning deep architectures for \{AI\}][Learning deep architectures for _AI],不过本文很长，很长……

5. Deep Learning工具——[Theano][]

[Theano][]是deep learning的Python库，要求首先熟悉Python语言和numpy，建议读者先看[Theano basic tutorial][]，然后按照*[Getting Started][]*下载相关数据并用gradient descent的方法进行学习。

学习了Theano的基本方法后，可以练习写以下几个算法：

有监督学习：

1.  [*Logistic Regression*][Logistic Regression] \- using Theano for something simple
2.  [*Multilayer perceptron*][Multilayer perceptron] \- introduction to layers
3.  [*Deep Convolutional Network*][Deep Convolutional Network] \- a simplified version of LeNet5

无监督学习：

*  [*Auto Encoders, Denoising Autoencoders*][Auto Encoders_ Denoising Autoencoders] \- description of autoencoders
 *  [*Stacked Denoising Auto-Encoders*][Stacked Denoising Auto-Encoders] \- easy steps into unsupervised pre-training for deep nets
 *  [*Restricted Boltzmann Machines*][Restricted Boltzmann Machines] \- single layer generative RBM model
 *  [*Deep Belief Networks*][Deep Belief Networks] \- unsupervised generative pre-training of stacked RBMs followed by supervised fine-tuning

最后呢，推荐给大家基本ML的书籍：

*  [Chris Bishop, “Pattern Recognition and Machine Learning”, 2007][Chris Bishop_ _Pattern Recognition and Machine Learning_ 2007]
 *  [Simon Haykin, “Neural Networks: a Comprehensive Foundation”, 2009 (3rd edition)][Simon Haykin_ _Neural Networks_ a Comprehensive Foundation_ 2009 _3rd edition]
 *  [Richard O. Duda, Peter E. Hart and David G. Stork, “Pattern Classification”, 2001 (2nd edition)][Richard O. Duda_ Peter E. Hart and David G. Stork_ _Pattern Classification_ 2001 _2nd edition]

关于Machine Learning更多的学习资料将继续更新，敬请关注本博客和新浪微博[Sophia\_qing][Sophia_qing]。

References：

1. [Brief Introduction to ML for AI][]

2.[Deep Learning Tutorial][]

3.[A tutorial on deep learning - Video][NIPS tutorial]

[1343977609_3572.png]: https://img-my.csdn.net/uploads/201208/03/1343977609_3572.png
[f458b0f65eb24e4d0f03630d0c18e607a61b0603.png]: /images/20220805/daa125824db04d7893e26a3b7da8c82f.png
[ab2c31dc00245ab18699df4af7ecfab088ef692e.png]: /images/20220805/7ee20bc4ac5649259706bddb25884abc.png
[Deep Learning]: http://www.iro.umontreal.ca/~pift6266/H10/notes/deepintro.html
[feature hierarchy]: http://blog.csdn.net/abcjennifer/article/details/7804962
[Link 1]: http://blog.csdn.net/abcjennifer/article/details/7748833
[3]: http://www.iro.umontreal.ca/~pift6266/H10/notes/deepintro.html#breakthrough-in-learning-deep-architectures
[1343714901_3624.png]: https://img-my.csdn.net/uploads/201207/31/1343714901_3624.png
[Learning Deep Architectures for AI]: http://www.iro.umontreal.ca/~lisa/publications2/index.php/publications/show/239
[webpage]: http://www.cs.toronto.edu/~rsalakhu/deeplearning/index.html
[list of references]: http://www.cs.toronto.edu/~rsalakhu/deeplearning/references.html
[public wiki]: http://www.iro.umontreal.ca/~lisa/twiki/bin/view.cgi/Public/WebHome
[reading list]: http://www.iro.umontreal.ca/~lisa/twiki/bin/view.cgi/Public/ReadingOnDeepNetworks
[bibliography]: http://www.iro.umontreal.ca/~lisa/twiki/bin/view.cgi/Public/DeepNetworksBibliography
[readings]: http://www.cs.toronto.edu/~hinton/deeprefs.html
[NIPS tutorial]: http://videolectures.net/jul09_hinton_deeplearn/
[A fast learning algorithm for deep belief nets]: http://www.cs.toronto.edu/~hinton/absps/fastnc.pdf
[Greedy Layer-Wise Training of Deep Networks]: http://www.iro.umontreal.ca/~lisa/publications2/index.php/publications/show/190
[Efficient Learning of Sparse Representations with an Energy-Based Model]: http://yann.lecun.com/exdb/publis/pdf/ranzato-06.pdf
[Learning deep architectures for _AI]: http://www.iro.umontreal.ca/~lisa/pointeurs/TR1312.pdf
[Theano]: http://deeplearning.net/software/theano
[Theano basic tutorial]: http://deeplearning.net/software/theano/tutorial
[Getting Started]: http://deeplearning.net/tutorial/gettingstarted.html#gettingstarted
[Logistic Regression]: http://deeplearning.net/tutorial/logreg.html#logreg
[Multilayer perceptron]: http://deeplearning.net/tutorial/mlp.html#mlp
[Deep Convolutional Network]: http://deeplearning.net/tutorial/lenet.html#lenet
[Auto Encoders_ Denoising Autoencoders]: http://deeplearning.net/tutorial/dA.html#daa
[Stacked Denoising Auto-Encoders]: http://deeplearning.net/tutorial/SdA.html#sda
[Restricted Boltzmann Machines]: http://deeplearning.net/tutorial/rbm.html#rbm
[Deep Belief Networks]: http://deeplearning.net/tutorial/DBN.html#dbn
[Chris Bishop_ _Pattern Recognition and Machine Learning_ 2007]: http://research.microsoft.com/en-us/um/people/cmbishop/prml/
[Simon Haykin_ _Neural Networks_ a Comprehensive Foundation_ 2009 _3rd edition]: http://books.google.ca/books?id=K7P36lKzI_QC&dq=simon+haykin+neural+networks+book&source=gbs_navlinks_s
[Richard O. Duda_ Peter E. Hart and David G. Stork_ _Pattern Classification_ 2001 _2nd edition]: http://www.rii.ricoh.com/~stork/DHS.html
[Sophia_qing]: http://weibo.com/u/2607574543
[Brief Introduction to ML for AI]: http://www.iro.umontreal.ca/~pift6266/H10/notes/mlintro.html
[Deep Learning Tutorial]: http://deeplearning.net/tutorial/