genann

Why we included this project

Genann is a good pick when you want a small feedforward neural network inside a C program without adopting a heavy framework. The whole library is two files, genann.c and genann.h, written in C99 with no dependencies, so it drops into embedded systems, games, or command-line tools and trains with backpropagation right away. The API stays minimal: init, train, run, read, write, and the weights live in one contiguous block, which makes it easy to swap in alternative training methods like hill climbing or genetic search. It ships with working examples for XOR and the IRIS dataset plus a test suite, so you can check behavior before trusting it. If you want a hackable, auditable network core rather than a large dependency, this is a practical fit.