Best free machine learning courses on YouTube, ranked by total hours
Updated 2026-09-22 · 9 courses · totals are live
Machine learning playlists range from a Stanford lecture hall to a whiteboard in someone's bedroom, and the bedroom is often the better teacher. This list mixes the two: full university courses (with the maths) and practitioner series (with the code). Every entry is a complete, free playlist you can follow from the first video to the last.
The hours are live totals from each playlist — the theory courses are longer than they look, and the maths-heavy ones don't speed-watch well.
From 3h 38m (Neural networks) to 3d 1h 19m (MIT 6.S191: Introduction to Deep Learning).
| # | Course | Videos | Total | At 1.5× |
|---|---|---|---|---|
| 1 | MIT 6.S191: Introduction to Deep Learning · MIT 6.S191: Introduction to Deep Learning — Alexander Amini MIT's fast, current deep-learning bootcamp, re-recorded each January. Neural nets to transformers and generative models. |
90 | 3d 1h 19m | 2d 53m |
| 2 | 100 Days of Machine Learning | CampusX · 100 Days of Machine Learning — CampusX (Hindi) A full practitioner's curriculum in Hindi: data cleaning, feature engineering, every classic model, with code. Very popular with Indian learners for good reason. |
134 | 2d 14h 50m | 1d 17h 54m |
| 3 | Complete Machine Learning playlist · Complete Machine Learning playlist — Krish Naik Broad, hands-on and interview-oriented. Good for filling gaps after a theory course. |
153 | 1d 12h 54m | 1d 36m |
| 4 | Machine Learning · Machine Learning — StatQuest with Josh Starmer Every core algorithm explained one idea at a time, with drawings. The best companion to any of the lecture courses. |
106 | 1d 5h 51m | 19h 54m |
| 5 | Stanford CS229: Machine Learning led by Andrew Ng | Autumn 2018 · Stanford CS229: Machine Learning — Andrew Ng (Autumn 2018) The canonical course. Linear models, generalisation, kernels, EM, RL — derived on the board. Expect to pause a lot. |
21 | 1d 3h 52m | 18h 35m |
| 6 | Stanford CS229: Machine Learning I Spring 2022 · Stanford CS229: Machine Learning (Spring 2022) The same course, more recent, taught by Tengyu Ma and Chris Ré. Slightly different emphasis; pick one, not both. |
19 | 1d 1h 43m | 17h 9m |
| 7 | Machine Learning with Python · Machine Learning with Python — sentdex Older, but it implements the algorithms from scratch in Python before using scikit-learn, which is the point. |
72 | 18h 48m | 12h 32m |
| 8 | Stanford EE104: Introduction to Machine Learning Full Course · Stanford EE104: Introduction to Machine Learning A gentler Stanford entry point (Sanjay Lall and Stephen Boyd): less proof, more intuition, still rigorous. |
19 | 14h 3m | 9h 22m |
| 9 | Neural networks · Neural networks — 3Blue1Brown Short and beautiful: what a neural network is, what gradient descent does, and how transformers work. Watch before anything else. |
10 | 3h 38m | 2h 25m |
Totals count available videos only and update as the playlists change. Private or deleted videos are excluded. Compare the top 5 side by side →
How to choose
- Want the maths? CS229 (either year) plus StatQuest for whenever a lecture loses you.
- Want to ship models? CampusX or Krish Naik, then MIT 6.S191 for deep learning.
- Only have a weekend? 3Blue1Brown's neural networks series. It's short, and it changes how the rest reads.
- Prerequisites — CS229 assumes linear algebra, probability and some calculus. If those are rusty, do CampusX's maths playlist or Khan Academy first.
Pace yourself
Theory lectures are the wrong place for 2× speed: the speed-watching guide has the evidence. Use the planner on each course page to schedule 45–60 minutes a day instead, and tick off lectures as you go.