This is a small pure C shared library for arbitrarily deep neural networks. It is my first attempt to write a scientific project in C but the speed is already outstanding. I made this library in order to assist me with my doctoral research.
- Python: python-libartificial
The feedforward procedure does not have a hardcoded depth (it can have as many layers as you want).
The library is created with Linux machines in mind but OSX users should not have a problem if they have gcc installed. I will try to compile the library with Visual Studio and get back to you on how to do it.
In order to get libartificial you have to do the following (assuming working installation of git)
git clone https://github.com/fetacore/libartificial.git cd libartificial rm -rf .git
Specifics for CLBlast
It is recommended to do the optimizations proposed by the author. The library assumes that CLBlast is in the libartificial folder under the name "clblast". You do the following:
git clone https://github.com/CNugteren/CLBlast.git clblast cd clblast && mkdir build && cd build cmake .. && make cd ../../
You should also take note that the library uses doubles and not all GPUs support double arithmetic in OpenCL.
In order to compile the library do the following (assuming you continue from where we left off)
- For CPU
- For GPU
For the time being I have four examples which you can find in the "examples" folder:
- MLP regression with CPU:
- MLP regression with GPU
- Autoencoder (CPU):
- CNN (only im2col at the moment):
This part is being written at the moment
If you like my work and/or you want to use it for your own projects or want me to create a custom recipe for you, I would gladly accept your donations at:
Copyright (c) Jim Karoukis. This project is licensed under the MIT License - see the LICENSE file for details.