About

Gorch

Gorch is a from-scratch automatic-differentiation and neural-network library written purely in NumPy. It was built as an educational project for a Neural Control course at K. N. Toosi University of Technology, with the goal of making backpropagation, optimizers, and network layers fully transparent — every backward pass is hand-written and verified against finite differences.

It pairs classic deep-learning machinery (tensors, autograd, 14 optimizers, modules and losses) with control-flavoured tools such as a KalmanFilter, an extended-Kalman neural identifier, and recursive least squares.

Author

Arman Gholibeikian — a machine-learning and control enthusiast who believes the best way to understand a tool is to rebuild it.

• GitHub: github.com/Armangb1
• Project: github.com/Armangb1/pygorch
• Email: arman.ghbn@gmail.com

License

MIT — see the repository’s LICENSE. Copyright (c) 2024 Arman Gholibeikian.