卷积神经网络小结(Convolutional Neural Networks)

CNNs应用的最成功的一个例子：Yann LeCun（曾经是Hinton组的research associate）

http://yann.lecun.com/exdb/lenet/index.html

code project上的一个C++写的例子，有很详细的文档说明：

http://www.codeproject.com/Articles/16650/Neural-Network-for-Recognition-of-Handwritten-Digi

代码下载网站：

MATLAB:

http://www.mathworks.com/matlabcentral/fileexchange/24291-cnn-convolutional-neural-network-class

https://sites.google.com/site/chumerin/projects/mycnn

C++:

http://www.codeproject.com/Articles/16650/Neural-Network-for-Recognition-of-Handwritten-Digi

文章进阶：

【1】Gradient-Based Learning Applied to Document Recognition

Yann Lecun写的文章。算是这个方向很重要的文章。

【2】Notes on Convolutional Neural Network：2006

主要是公式推导。

【3】Best Practices for Convolutional Neural Networks Applied to Visual Document Analysis:2003

对于结构把握比较好，如果要自己去编写代码，可以仔细看看这篇文章。

【4】3D Convolution Neural networks for Human Action Recognition:2012

以前的CNNs都是2D的，But this paper is for human action recognition, so one has to apply temporal dimension. So the author developed a noval model:3D convolution neural networks.(In 2D CNNs, 2D convolution neural networks is performed at the convolution layers to extract
the feature from local neighbourhood on feature maps in the previous layer.)