gradient-descent-edf1101

Gradient Descent Module

Introduction

Gradient Descent for the Fundamentals of Machine Learning module of my Computer Science Degree

Getting Started

Installation

This can be installed via pip by running

pip install

It can then be used as a regular module

Examples

Examples found in the examples folder both as a python file and as jupyter notebooks

• The blue points are the random points that somewhat fit the curve 2x² - 3x + 4
• The Red line is the actual curve
• The Green line is the predicted curve

Prerequisites

Modules Used

• Matplotlib

Version of python Used = 3.11.4

Developer Details

Documentation

HTML file documentation generated by sphinx can be found here

Linting

In the base of the project you can run pylint gradient_descent

Learning Parameters (for OK results)

• Quartic - learning = 0.0000000003
• Cubic - learning = 0.00000003
• Quadratic - learning = 0.000001
• Linear - learning = 0.00001

In general for each extra polynomial term you add it should be ~100x smaller. Smaller learning rates may get better results, but you will need to have more attempts to reach it since it learns slower

Testing

pytest is required for these tests. This can be installed by pip install pytest

Details

License

This falls under the MIT license found here

Authors

Ed Fillingham

Sources

This project can be found at https://github.com/edf1101/GradientDescent