SIMPLEST Neural Network Possible

This is THE SIMPLEST neural network implementation you will find (...that's still useful.)

The neural network uses your GPU to train on the MNIST dataset and learn to recognize images of hand-written digits.

If you don't have CuPy installed on your system, the program will use NumPy (CPU) instead. The training will run much slower, but it will achieve the same results.

"simplest-nn-oo.py" is just an object oriented version of "simplest-nn.py"

They do the same thing and achieve the same results. The OO version is just there in case you want to see how to implement it in that way.

Email me@johnathanalyst.com with questions.

Video Demo in my AI Finance series

Enjoy!

Prerequisites

• Python
• NumPy
• CuPy (if you want to use your GPU)
• Requests (unless you already have the csv files: train & test)
• Matplotlib
  
  

Installation

 git clone https://github.com/chivington/SIMPLEST-Neural-Network-Possible.git

Usage

cd SIMPLEST-Neural-Network-Possible/data
python download.py
cd ..
python simplest-nn.py
python simplest-nn-oo.py

For more details, see the Video Demo Feel free to ask me questions on GitHub

Authors

• Johnathan Chivington: GitHub

Contributing

Not currently accepting outside contributors, but feel free to use as you wish.

License

There is no license associated with this content.