论文笔记:Going deeper with convolution

最新推荐文章于 2023-01-28 11:21:36 发布

John2King

最新推荐文章于 2023-01-28 11:21:36 发布

阅读量671

点赞数

CC 4.0 BY-SA版权

分类专栏： DL 文章标签： cnn googlenet

本文链接：https://blog.youkuaiyun.com/lebula/article/details/51699122

DL 专栏收录该内容

15 篇文章

订阅专栏

本文探讨了Inception网络设计背后的动机：如何使卷积神经网络既深又宽的同时避免资源限制和过拟合问题。文章提出了两个解决方案：一是通过1x1卷积减少冗余信息并压缩特征；二是引入辅助损失函数加速训练过程。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

1. Motivation of inception: To make deeper and wider CNN comes with two drawbacks: limitation of examples and increasing computational resources. ==> need to move from fully connected to sparsely connected architectures.

2. Solutions:

i. use 1 x 1 convolutions before 3 x3 and 5 x 5 ==> reduce redundant infor and reduce dim ( 'keep the network sparse at most places and compress the signals only whenever they have to be aggregated')