Striving for Simplicity: The All Convolutional Net

http://arxiv.org/abs/1412.6806

CNN网络貌似由简单到复杂，再由复杂到简单的趋势

当前主流的CNN网络一般都包括以下几个模块：convolution and max-pooling layers和交替，最后是少数全链接层。本文对此深入分析，提出了只有卷积层的网络，发现效果不错。

首先来看看 pooling， why pooling can help in CNNs，可能的解释有三个：
1） the p-norm makes the representation in a CNN more invariant
2） the spatial dimensionality reduction performed by pooling makes
covering larger parts of the input in higher layers possible
3） the feature-wise nature of the pooling operation (as opposed to a convolutional layer where features get mixed) could make optimization easier.

假定第二条是最重要的，那么我们可以有2中方法得到去除pooling但保留spatial dimensionality reduction
1）We can remove each pooling layer and increase the stride of the convolutional layer that preceded it accordingly.
2）We can replace the pooling layer by a normal convolution with stride larger than one。（对于一个k=3,r=2的池化层，我们可以用一个对应的stride和kernel的卷积层取代，输入个数输出个数一样。）

第一个方法有一个缺陷： we significantly reduce the overlap of the convolutional layer that preceded the pooling layer. It is equivalent to a pooling operation in which only the top-left feature response is considered and can result in less accurate recognition.

第二个方法则没有这个问题。

对于全链接层， if the image area covered by units in the topmost convolutional layer covers a portion of the image large enough to recognize its content (i.e. the object we want to recognize) then fully connected layers can also be replaced by simple 1-by-1 convolutions.

Striving for Simplicity: The All Convolutional Net相关推荐

【论文笔记 2】CNN经典入门STRIVING FOR SIMPLICITY: THE ALL CONVOLUTIONAL NET
[论文笔记 2]CNN经典入门STRIVING FOR SIMPLICITY: THE ALL CONVOLUTIONAL NET 1. 综述对现代的CNN网络结构进行了简要说明,一般的网络都是由卷 ...
[深度学习论文笔记][Visualizing] Striving for Simplicity: The All Convolutional Net
Springenberg, Jost Tobias, et al. "Striving for simlicity: The all convolutional net." arX ...
《STRIVING FOR SIMPLICITY: THE ALL CONVOLUTIONAL NET》论文学习
ABSTRCAT 如今用于目标识别的CNNs大都以相同的原则构建:卷积层与最大池化层交替,接着是少量全连接层.本文重新评估了用CNN从小图像中识别物体的技术水平,产生了对上述结构中不同成分的必要性的质 ...
深度学习的可解释性——Striving For Simplicity: The All Convolution Net
Striving For Simplicity: The All Convolution Net 是ICLR 2015的一篇论文,作者Jost Tobias Springenberg , Alexey ...
Striving For Simplicity: The All Convolution Net
Striving For Simplicity: The All Convolution Net 是ICLR 2015的一篇论文,作者Jost Tobias Springenberg , Alexey ...
Striving For Simplicity: The All Convolution Net 简析
转载[https://blog.csdn.net/qinqbaobei/article/details/54092619] Striving For Simplicity: The All Convo ...
卷积神经网络（Convolutional Neural Networks，CNNS/ConvNets）
本文翻译自 Convolutional Neural Networks(CNNs / ConvNets),更多内容请访问:http://cs231n.github.io/. 原来译文:https:// ...
【卷积神经网络】卷积神经网络（Convolutional Neural Networks, CNN）基础
卷积神经网络(Convolutional Neural Networks, CNN),是一种针对图像的特殊的神经网络. 卷积神经网络概述 Why not DNN? 图像数据的维数很高,比如 1, ...
Convolutional Neural Networks for Visual Recognition
http://cs231n.github.io/ 里面有很多相当好的文章 http://cs231n.github.io/convolutional-networks/ Table of Cont ...

Striving for Simplicity: The All Convolutional Net

Striving for Simplicity: The All Convolutional Net相关推荐

最新文章

热门文章