HuyNguyenAu / wobbly-ml

An ML library based on the Neural Networks From Scratch book.

Wobbly ML

A ML library based on the Neural Networks From Scratch book.

Running the example regression gives the following results:

We can also see that the loss is very noisy. Not every impressive, but it works! Yay?

Requirements

• Python 3.9.13 minimum.
• Python Libraries:
  • numpy
• Libraries For Examples:
  • nnfs
  • cv2
  • matplotlib

Setup

Install the following libraries:

$ pip install numpy

Install wobbly-ml as a dev library:

$ sh install.sh

Running the examples will require additional libraries:

$ pip install matplotlib nnfs opencv-contrib-python-headless

Usage

Run the example regression.py file:

$ python3 regression.py

Testing

Errr... No tests yet.

Why?

The goal of this project was to try to understand what a neural network is and how it is built. I wanted to build this simple and small library because I did not agree with how the book structured the project. That was a good exercise because I had to think careful about where each functionality belonged to.

Why WobblyML?

I don't expected this to be stable, production ready, or accurate.

Let's just have some fun and explore!