schikani/nn ? reverse-engineered prompt

Build me a simple neural network project that can train and test on data from CSV files, with a small example app I can run from the command line.

I want it to be easy to compile, with a clean CMake build, and to include a basic model implementation, some reusable headers, and a few sample datasets or example inputs so I can try it quickly. If there are missing pieces, please fill them in and make the project feel complete, with clear defaults and a straightforward way to run a demo.

It should be okay if you need to look up current docs online for anything uncertain. Please keep the code organized and readable, and make sure the example actually shows the neural network working end to end.