Big Brian

"Sometimes my genius, it's... it's almost frightening." - Jeremy Clarkson

When I initially created this project, I actually just misspelled "brain", so yeah.

This project is a playground for me to learn more about neural networks and machine learning.

Goals

• Convolutional neural network with hidden layers and per-node biases.
• Propagation of a number of a population of networks with a set of data, plus scoring and ranking for the population.
• Network training using back propagation.
• WASM/JS/TS package and sample website?

Usage

Requirements

This project uses CMake version ^3.20 (literally just picked a version) and C++14.

Building

To build, run the following commands:

# Create a build directory
mkdir build && cd build

# CMake to generate build files
cmake ../

# Build the project (you can replace this with whatever process is needed for the generator you used)
cmake --build .

My default on my PC was set to "Visual Studio 17 2022" and it failed for some unknown reason when using the build command, but I normally do "MinGW Makefiles", and that worked so I don't know what's up with that.

Testing

There are unit tests setup with CTest, which can be run with the following:

ctest

Running

Currently, there is no main program to use.

Previous Iteration

This project was originally written in C#, implementing an convolutional neural network, utilizing back propagation for training.

It was able to "read" hand-written numbers from the MNIST database.

License

Copyright (c) 2024 Hunter Barclay

MIT License.