Visualizing and Understanding convolutional networks

Large convolutional networks model on ImageNet 

(Krizhevsky, A., Sutskever, I. and Hinton, G. E. ImageNet Classification with Deep Convolutional Neural Networks , Advances in Neural Information Processing 25, 2012)

Why they perform so well?

How they migh be improved?


为什么CNN 在imagenet上取得如何显著的结果?可以归结为下面三个当面:

1、larger training sets;  millions of labeled examples. 

2、GPU makes practical

3、Better model regularization strategies: Dropout


输入:2D images;  输出:a probability vector over the C different classes.

Convolutional layers 包含:

(1) Convolution of the previous layer output with a set of learned filters

(2) Passing the responses through a rectified linear function (relu(x) = max(x,0));   Ensure the feature maps are always positive.

(3) Max-pooling over local neighborhoods (optionally) 

(4) Normalize the responses across feature maps.(optionally) 

Top few layers of NN: Fully-connected networks

Final layer: a softmax classifier.


Visualization with a Deconvnet by mapping these activities back to the input pixel space. 


评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值