博客
关于我
强烈建议你试试无所不能的chatGPT,快点击我
Convolution 中 group 参数
阅读量:5881 次
发布时间:2019-06-19

本文共 1109 字,大约阅读时间需要 3 分钟。

  1. caffe 官方文档定义:

    group (g) [default 1]: If g > 1, we restrict the connectivity of each filter to a subset of the input. Specifically, the input and output channels are separated into g groups, and the i-th output group channels will be only connected to the i-th input group channels.

  2. Jia Yangqing 论坛回复:

    It was there to implement the grouped convolution in Alex Krizhevsky's paper: when group=2, the first half of the filters are only connected to the first half of the input channels, and the second half only connected to the second half.

    To group seems like a historical accident from the practical constraints of the time, but I don't know of a side-by-side comparison of group and no-group models. It could be nice to settle the question on the axes of task performance, speed + memory, and optimization time + hassle.

  3. Alex Krizhevsky's paper 中conv的定义:

    convolution_param {

    num_output: 256
    pad: 2
    kernel_size: 5
    group: 2
    weight_filler {
    type: "gaussian"
    std: 0.01
    }
    bias_filler {
    type: "constant"
    value: 1
    }

结论:Alex文章中是多个显卡,可能基于并行内存等考虑。所以以后少用。。。

reference:

转载于:https://www.cnblogs.com/penguins/p/5103360.html

你可能感兴趣的文章
PyTorch快速入门教程六(使用LSTM做图片分类)
查看>>
什么是go语言
查看>>
Java中文排序完美新解
查看>>
Hadoop jps正确使用流程及报错处理
查看>>
docker learn
查看>>
pexpect 'module' object has no attribute 'spawn'
查看>>
手机自动化测试:appium源码分析之bootstrap九 1
查看>>
linux学习工作中遇到的问题及解决办法
查看>>
分享10个帮助你实现超酷特效的jQuery插件
查看>>
atm仿真脚本
查看>>
11套免费的wordpress模板主题
查看>>
Facebook 新闻(Facebook Journalism)
查看>>
解除apt-get被占用
查看>>
PHP Smarty 模板 读取配置文件变量
查看>>
springboot集成mqtt
查看>>
Java 数组 之 一维数组 删除 元素
查看>>
【学习笔记】Android控件的使用
查看>>
mac 下载软件无法将程序拖移至app 出现错误
查看>>
关于int(1)和int(10)的区别
查看>>
动态域名解析制作
查看>>