Implement artificial neural network for MNIST classification task using basic C programming language.
Get my hands dirty and learning deep learning the hard way.
- For Chinese students-"这是一份可以当作 C 语言大作业的项目"
- Modified MNIST dataset which can be downloaded from kaggle.
This is a demo project you can modify main.c to customize the program. In windows, you can either compile with cl main.c simplenet.c simplela.c datareader.c command or add source files to an visual studio project, in linux, use gcc to replace cl for compiling.
You will get an executable file named main.exe or some thing like that, run it by click or call in command line as you like.
Completed:
- Refactoring and saved myself from annoying pointer operations
- Simple linear algebra structure like vector and metrix, operations like add and multiply
- Basic net structure with
structattribute in C - Forward pass in simple feedforward fully connected network
- Backward pass in simple feedforward fully connected network with mini-batch sized 1
In progress:
- Backward pass with full batch
Plan:
- Introduce tools like glog to help with debugging and rewrite test files
- Write makefile for the project and support compling in Windows and Linux