The "Mathematics for Machine Learning" book — what it is, who it's for, and how to get ready
A short, honest guide for anyone who found this page searching for the book, or its PDF.
If you've been pointed toward machine learning and keep hearing the same title come up, you've probably found Mathematics for Machine Learning by Marc Peter Deisenroth, A. Aldo Faisal and Cheng Soon Ong (Cambridge University Press). It's widely considered the standard rigorous reference for the maths underneath ML — clear, well-organised, and free to read. The authors publish the official PDF themselves at mml-book.github.io, which is the right place to get it.
Mathematics for Machine Learning (math4ml.co) is an independent course and is not affiliated with, endorsed by, or connected to the authors of the mml-book or Cambridge University Press. We simply think it's an excellent resource and want to help you get the most out of it.
What the book actually is
The mml-book is a full university-style textbook. It covers linear algebra, analytic geometry, matrix decompositions, vector calculus, probability and continuous optimisation in Part I, then applies all of it to real ML models — linear regression, PCA, Gaussian mixture models, SVMs — in Part II. It's dense, precise, and written the way a graduate maths course would be written: definitions, theorems, proofs, worked derivations. That rigor is exactly why it's so respected, and exactly why it can be a hard place to *start*.
In the authors' own preface, they're upfront that the book is aimed at readers who already have a university-level mathematical background — it is not designed to teach calculus, linear algebra or probability from scratch. It assumes you can already read and manipulate the notation, not just recognise the words.
You're probably ready for it if…
- You can differentiate and integrate standard functions without looking up the rules.
- Matrix multiplication, determinants, eigenvalues and vector spaces are familiar territory, not new vocabulary.
- You're comfortable reading summation (Σ), product (∏) and set notation at a glance.
- Probability distributions, expectation and variance make sense to you algebraically, not just intuitively.
- You've taken (and retained) at least one university-level course covering some of the above.
If most of that sounds familiar, the mml-book is a genuinely excellent next step — dive in.
If that list felt shaky: start from zero here first
If the notation itself is the obstacle rather than the ML ideas, that's completely normal — it's the single biggest reason people bounce off rigorous math texts. Mathematics for Machine Learning (math4ml.co) is built for exactly that gap: it starts from zero, with no assumed background, and builds the same foundations — algebra, linear algebra, calculus, probability and statistics — through plain-language explanations, interactive draggable figures you can move with your own hands to build intuition, and unlimited auto-graded practice so the ideas actually stick before you meet them in denser notation. It runs entirely in your browser, in 19 languages.
The Foundations course (24 lessons) is free forever — no card, no account needed to start. You can also try two full sample lessons from deeper in the course, free, right now:
- The Gradient — from Calculus II
- Eigenvectors & Eigenvalues — from Linear Algebra
Using them together
You don't have to pick one. A natural way to combine them: work through math4ml's Foundations and Linear Algebra / Calculus courses to get fluent with the notation and build real intuition, then read the mml-book chapter by chapter as your rigorous reference and deep dive — using our interactive practice to warm up on a topic (say, eigendecomposition or gradients) right before you read the matching mml-book chapter. The book will go further and deeper than any from-zero course can; the point of math4ml is just to make sure you're not decoding notation and learning the ML idea at the same time.