Math for ML Video Courses: Weights & Biases' Math4ML and 3Blue1Brown
If you're searching for "math for machine learning video course" or "wandb math4ml," you've probably already found two of the best free resources on the internet for building mathematical intuition for ML. This page is a short, honest guide to both — what they're great for, who they assume you already are, and what to do if you're not quite there yet.
What these video courses are — and who they're for
The Weights & Biases Math4ML series, taught by Charles Frye, is a set of free lecture-style videos on the Weights & Biases YouTube channel, covering the calculus, linear algebra, probability and optimization that underpin machine learning — aimed squarely at the maths you actually need for ML papers and models, not maths for its own sake. W&B also maintains a companion exercises repository on GitHub with problem sets to work through alongside the lectures.
3Blue1Brown's Essence of Linear Algebra, at 3blue1brown.com/topics/linear-algebra, is a different kind of resource: a short, beautifully animated series that rebuilds your geometric intuition for vectors, matrices, determinants and eigenvectors from first principles. It's widely considered one of the best explanations of linear algebra ever made, free or paid.
Both are made by people who are exceptionally good at explaining ideas visually, and both are genuinely worth your time. Math4ML is pitched at someone who wants the specific maths of ML lectures-first, with exercises alongside; Essence of Linear Algebra is pitched at anyone who wants deep, lasting geometric intuition for one specific subject, in a very tight, rewatchable format.
Mathematics for Machine Learning (math4ml.co) is an independent project and is not affiliated with, endorsed by, or connected to Weights & Biases, Charles Frye, or 3Blue1Brown (Grant Sanderson) in any way.
You're ready for them if…
- You're comfortable with basic algebra — manipulating equations, functions, exponents — without having to think hard about it.
- You've seen derivatives and integrals before, even if they're rusty, and the words "chain rule" or "partial derivative" don't feel completely foreign.
- You know what a vector and a matrix are, and ideally have multiplied a couple of matrices by hand at some point.
- You're comfortable learning by watching — pausing, rewinding, and working through examples on paper yourself — rather than needing built-in graded practice to check you've actually got it.
- You want intuition and the "why" behind ML maths more than step-by-step drilling.
If you're not quite ready: start from zero here
If any of that checklist felt shaky, that's completely normal — and it's exactly the gap Mathematics for Machine Learning (math4ml.co) is built to close. It's a from-zero, interactive course: 151 lessons across 6 courses, each with a plain-language explanation, a draggable figure you can manipulate to build intuition, and unlimited auto-graded practice problems that tell you immediately whether you've actually understood something — not just watched it.
The Foundations course (24 lessons) is free forever and starts from the absolute basics — no calculus or linear algebra assumed. You can also try two full sample lessons with no sign-up at all: The Gradient and Eigenvectors & Eigenvalues — the second one pairs especially well with 3Blue1Brown's eigenvector video, since you can watch the intuition there and then immediately test it with an interactive, auto-graded lesson here.
Using them together
These aren't competing options — they cover different needs and work well side by side. A practical way to combine them: watch the relevant 3Blue1Brown or Math4ML video first for the big-picture intuition and the "aha," then come to Mathematics for Machine Learning and work the matching topic's interactive lesson and practice problems until the mechanics are second nature, not just something you remember watching. If a video assumes something you haven't covered yet — say, partial derivatives before you've seen ordinary ones — use the from-zero Foundations lessons here to fill that specific gap, then go back to the video with the prerequisite in place. Video-first for intuition, practice-first for retention is a combination that works well for most people learning ML maths from scratch.