// the find
Sentdex/nnfs_book
Sample code from the Neural Networks from Scratch book.
Companion code for the Neural Networks from Scratch book by Harrison Kinsley (Sentdex), which walks through building neural networks using only NumPy — no PyTorch, no TensorFlow. It's aimed at people who want to understand what's actually happening inside a network before reaching for a framework.
Each chapter is a single self-contained Python file, which makes it easy to follow along or jump to a specific concept. The NumPy-only approach means you're forced to implement backprop, weight updates, and activation functions yourself — that friction is the point. The companion YouTube series (Sentdex's channel) adds video context that the code alone doesn't provide. The nnfsix repo extending this to 40+ languages shows the core code is clean enough to port.
One final file per chapter with no intermediate steps means you can't see how the code evolved during the chapter — you get the answer but not the journey. The repo hasn't been touched since mid-2024 and chapters 7, 8, 12, 13, 20, and 21 are simply missing, presumably covered only in the paid book. Zero tests or assertions, so if you tweak something and break it, you're debugging by inspection. This is reference material, not a library — there's nothing to install or run end-to-end.