Rewrite the Deep Learning from Scratch book code in F# Rust. Eventually write some of the portions in WebGPU for speeding up. This is something worth learning.
The library will be written in Rust for a couple of reasons.
- To be in a separate language from Python, which this book is written in so that no copy/paste is possible
- Rust seems like a good language for performance and has good libraries for GPU, dataframes, and visualizations
Chapter 1 - Math Library
Chain Rule #next
Multi Variable Calculus
Vector Calculus
Chapter 2 - Neural Network
Perceptron
Single Layer
Multi-layer network
Perhaps using the PyTorch API