ML-From-Scratch

Plain Python and NumPy code that implements classic machine learning models so you can read how they work.

Stars: 33.3k
Forks: 5.5k
Licence: MIT
Language: Python
Last update: 3.1 years ago
Type: Learning resource

No commits in over a year — check for forks or alternatives before adopting.

What it is

This project contains simple, from-scratch Python implementations of fundamental machine learning models and algorithms, built on NumPy. Its stated purpose is not speed or optimization but showing the inner workings of each method in a transparent, accessible way. The README says it aims to cover topics from linear regression to deep learning, and it includes runnable examples such as polynomial regression, a CNN digit classifier, DBSCAN clustering, and a GAN.

Run the polynomial regression example to watch a model fit temperature data
Run the DBSCAN example to see density-based clustering on the moons dataset
Run the genetic algorithm example to see a target string evolved from random candidates
Read the CNN or DQN code to understand how layers and reinforcement learning agents are built

Great for

Students who want to see how algorithms like regression or neural networks work internally
Developers who learn better by reading code than by calling a library
People studying machine learning who want to step through the math in code
Teachers preparing examples that show algorithm internals

Not ideal for

Production systems that need speed or scale, because the README says efficiency is not the goal
People who only want a ready-made library to call with minimal code
Beginners with no Python experience, since the examples assume some comfort with the language and ML concepts

Step-by-step guide: how to use ML-From-ScratchInstall, first run, examples, tips and FAQs

How to get it

1
Clone and build from source
You'll need: Python · Git
Clone the repository
git clone https://github.com/eriklindernoren/ML-From-Scratch
Enter the project folder
cd ML-From-Scratch
Install the package
python setup.py install
2
Run it
- Clone the repository and change into its folder
- Run the install command from the README
- Run an example such as python mlfromscratch/examples/polynomial_regression.py
- Read the matching source file under mlfromscratch to see how the model is implemented
3
Get the most out of it
- Start with the polynomial regression example, since it is the first one in the README and is the simplest to follow
- Compare the printed layer tables in the CNN and GAN examples with the code that builds them
- Modify the genetic algorithm target string parameter to see how the search behaves
- Use the examples as a reference when reading the Implementations section of the README

Downloads

Source code (.zip)master branch · straight from GitHub

Strengths

Written in plain Python and NumPy, so the logic is visible rather than hidden in a framework
Includes several runnable examples with printed model summaries and training output
Released under the MIT license

Trade-offs

The README states it is not designed for optimized or efficient computation
The README excerpt provided gives limited detail on installation requirements and which parts are complete
Last push in the metadata is October 2023, so it may not reflect recent changes

How it compares

Project	Stars	Type	Level	Updated	Best for
ML-From-Scratch (this)	33.3k	Learning resource	beginner	3.1 years ago	Students who want to see how algorithms like regression or neural networks work internally
mit-deep-learning	10.5k	Learning resource	beginner	2.8 years ago	Students following MIT deep learning courses
awesome-datascience	30.3k	Curated list	beginner	6 days ago	Beginners who want a guided set of links to start learning data science
Made-With-ML	49.9k	Learning resource	intermediate	8 months ago	Software engineers who want to learn how ML models get deployed and maintained
python-machine-learning-book	12.7k	Learning resource	beginner	4 months ago	Readers who own the 1st edition and want to run its code
AI-Expert-Roadmap	31.3k	Learning resource	beginner	1.2 years ago	Beginners who are unsure what to learn next in AI or data science

Stats synced from GitHub. The assessment above was drafted with AI from the project's README on 2026-10-29 and may miss recent changes — check the README before relying on a command. MIT licence: free to use, modify and distribute, including commercially, as long as you keep the copyright notice.