Skip to content

Latest commit

 

History

History
16 lines (7 loc) · 741 Bytes

File metadata and controls

16 lines (7 loc) · 741 Bytes

Automatic Differentiation from Scratch in Python

This repository contains an extremely basic machine learning framework with automatic differentiation in a couple lines of plain Python.

With inspiration from [Baydin2018].

There's no support for vectorization or GPUs. Performance will be terrible.

See the tests for api usage, and demo.py for a more complex example.

[Baydin2018]Baydin, Atilim Gunes, Barak A. Pearlmutter, Alexey Andreyevich Radul, and Jeffrey Mark Siskind. 2018. “Automatic Differentiation in Machine Learning: A Survey.” Journal of Machine Learning Research 18 (153): 1–43. https://www.jmlr.org/papers/volume18/17-468/17-468.pdf