Skip to content

Inplement an CNN frame with Numpy, easy to learn, hard to use hhhh

Notifications You must be signed in to change notification settings

AirBernard/CNN-Numpy

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

6 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

CNN Inplementation With Numpy

​ 复习深度学习相关知识。打算用numpy实现一个简单的可以训练,测试的cnn框架(model-free,model-based),包含一些主流的层和设计,以便复习与巩固基础。

2018.01.22


Target1: (model-free)实现mnist的训练与测试:

  • layer: Conv2D, FullyConnect, MaxPooling, Softmax
  • activation: Relu
  • method: Mini-batch Gradient Descent(SGD),learning_rate = 1e-5

version validation_acc train_acc inferencetime(ms/pf)
baseline 96.75% 97.15% 2(ms/pf)

2018.01.24


Target2:  (model-based)实现Variable与Operator分离设计:

  • 完成Variable与Operator 类的设计与graph的注册功能,GLOBAL_VARIABLE_SCOPE作为全局所有Variable,Operator的索引(graph),Operator,Variable类自己维护自己的child,parent列表。(感觉有点像tf)
  • 完成Conv2D类的设计,对比上一版本进行测试通过。

2018.01.25


  • 完成其他基本组件的Operator改写,新版本的mnist训练。

About

Inplement an CNN frame with Numpy, easy to learn, hard to use hhhh

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%