Stanford University (cs229.stanford.edu, YouTube StanfordOnline)
Stanford CS229 Machine Learning Review — Honest Analysis of Andrew Ng's Original Course
Stanford CS229 is the original Andrew Ng university course — math-heavy, blackboard-driven, ~20 lectures of full derivations free on YouTube and cs229.stanford.edu. It is the polar opposite of the new Coursera Machine Learning Specialization in tone and audience. Take it if you want to understand why the algorithms work, not just how to call them. Skip it if you are a beginner without comfortable linear algebra and probability — by the consistent reviewer testimony, that wall is real.
Final score
from 32 analysed opinions
Published AI-researched, editor-audited
Distribution of opinions
Per-criterion scores
Reviewers consistently praise the mathematical depth — full derivations of GLMs, SVMs, EM, factor analysis and learning theory. The honest caveat is that the curriculum predates the Transformer era and deep learning gets brief treatment.
Andrew Ng's blackboard teaching gets repeated praise — one HN reviewer specifically prefers it to the Coursera version because he uses the board. The lecture pacing is academic and unhurried, which some find rigorous and others find slow.
Completely free — full 2018 lecture series on YouTube, all lecture notes, problem sets and section materials at cs229.stanford.edu. No certificate, no grading, no paywall. Reviewers consistently call it the highest-value rigorous ML resource available.
Zero official support for the YouTube cohort — no forum, no grading, no TA office hours, no cs50.ai-style tutor. Self-learners rely on community GitHub repos for solutions. Honest weakness, not unique to CS229.
Theory transfers durably — gradient descent, GLMs, regularisation, EM and learning theory remain foundational. The honest gap is that CS229 was not designed as a practical-first course; deployment, modern frameworks and Transformers are out of scope.
What learners said
What people loved
5- Mathematical rigour — full derivations of GLMs, SVMs, EM, factor analysis, learning theory×14
- Andrew Ng's blackboard teaching is preferred over the slide-driven Coursera version×8
- Completely free — lectures on YouTube, notes and problem sets at cs229.stanford.edu×11
- Lecture notes are dense enough to function as a textbook substitute×7
- Covers theory other courses skip — VC dimension, learning theory, reinforcement learning×6
What frustrated learners
5- Math prerequisite is real and underestimated — linear algebra, probability, multivariable calculus×9
- Curriculum predates Transformers and modern deep learning; CNN and LLM coverage is minimal×7
- No support, no grading, no forum for self-learners watching on YouTube×5
- Pacing is academic and lectures are long — not designed for time-constrained learners×4
- Practical and applied content is thin — this is a theory course, not a path to shipping models×4
Real quotes from real users
“I took CS229 here at Stanford and I was also one of the TAs for the online version last year. The Stanford CS229 version is definitely much more difficult than what you guys had online. We had to take it down a notch and water it down to take it online.”
“I have found Andrew Ng's math handouts for CS229 Stanford (not Coursera) the best. There's necessary introduction and abstraction of irrelevant details to the topic at hand.”
“Been interested in ML lately. I'm a good half way through Andrew Ng's CS 229 (the actual course, not the watered down coursera course). I find it much more intellectual stimulating than writing apps.”
“For those with a strong mathematical background I would recommend Stanford's CS229 with Andrew Ng. I think it's better than the Coursera version because he uses the board.”
“At one point in time I also went through half of Andrew Ng's CS229 ML lectures from Stanford. Found them much more math/proof heavy, but definitely valuable for understanding the underlying statistical methods and theory that ML apply.”
“I have an unpopular opinion — I didn't like CS229 that much. The course is ambitious. The lecture notes are dense. The assignments are difficult, but still nowhere as difficult as the real exam. The median the year I took it was 46/100.”
“Andrew Ng's CS229 course on machine learning has earned a reputation as one of the best resources for learning the fundamentals. It is a demanding class that requires a significant amount of time and effort and a strong foundation in linear algebra and good understanding of probability and statistics.”
“When I did the Coursera course I supplemented it with the CS229 materials to get more of the theory behind everything, which worked great. If you want more math rather than less, it's the way to go.”
Frequently asked questions
How we evaluated this
This review synthesizes 32 opinions collected across the public web. Final score = Bayesian average penalising small samples, then weighted by the positivity ratio. No paid placements, no hidden agenda.
- 26 from Hacker News
- 6 from Blogs