This project implements a vanilla neural network in c++, which is useful for some types of classification. By vanilla, I mean not deep or complicated. There's one activation function.

I'm looking back on this project now, six months after it began, and regretting my decisions. I could restructure this but I don't have the time. It's not all bad as a first c++ project. Lesson learned: Dependency Injection is important and it's better to get it right the first time rather than fixing it later.

Docs

• using the program
• data file structure
• example
• issues with the program
• reasons for project structure
• optimistic list of things to add

Compile

On a Linux system, do: make at your terminal.

This project was made on a Linux system, and there's no plan to make sure it works in Windows or Mac. Make a pull request if that's what you want.

Tests

If you're missing permissions, do: chmod +x bin/*

Then do ./bin/tests. Most of these tests are integration tests.