Back to index
OtherSelf paced, a large free library of short focused videos on data science, statistics and machine learning maths·Free

ritvikmath (YouTube)

4.3

The channel to reach for when the maths under a machine learning method is the part that is losing you. Ritvik has a real gift for taking a topic like Bayesian inference, time series or the bias variance trade off and explaining it in ten focused minutes with genuine intuition rather than hand waving.

What We Liked

  • Exceptionally clear at explaining the statistics and probability that most ML courses rush past
  • Short, single topic videos make it easy to find and fix one specific gap in your understanding
  • Whiteboard style teaching that builds intuition rather than just stating formulas
  • Completely free and covers topics, like Bayesian methods and time series, that are underserved elsewhere

What Could Be Better

  • It is a topic by topic library, not a structured course, so you have to build your own path
  • Very little hands on coding, so you learn the why far more than the how in practice
  • Best suited to someone who already knows what concept they are missing rather than a total beginner
  • Depth is capped by the short format, so genuinely advanced treatment needs heavier resources

Detailed review

Ritvik Kharkar's channel, ritvikmath, occupies a specific and genuinely valuable niche, the mathematics and statistics that sit underneath data science and machine learning and that so many courses skate over on the way to the code. A lot of learners hit the same wall, they can import a library and fit a model but they do not really understand the probability, the distributions, the inference or the assumptions underneath, and this is the channel that tends to get them over it. Ritvik works in short, single topic videos, usually with a whiteboard or a tablet, and his strength is intuition, so instead of putting a formula on screen and moving on he builds up why the formula looks the way it does and what it is really doing. That makes topics that are often taught badly, Bayesian inference, maximum likelihood, the bias variance trade off, time series models, hypothesis testing, land in a way they frequently do not elsewhere, and he covers some corners, particularly Bayesian methods and time series, that are noticeably underserved by the big video courses.

Because each video is self contained and clearly titled, the channel doubles as a kind of searchable reference, so when a specific idea in your main course will not click you can very often find a ten minute explanation here that fixes it. The trade offs are the natural consequences of the format. This is a library rather than a curriculum, so there is no built in order and no sense of progression, and you are responsible for knowing what you need and stitching the pieces together yourself. It is also deliberately light on code, so it teaches you to understand methods far more than to implement them, which means it pairs with a hands on course rather than replacing one.

The same short format that makes it so approachable caps how deep any single video can go, so once you are past intuition and into genuinely advanced treatment you will need to move to books or longer courses. And a complete beginner who does not yet know which concept is missing will get less from it than someone who can already point at their specific gap. My take is that ritvikmath is one of the best free resources going for the mathematical underpinnings of machine learning, as long as you use it as a supplement. Learn from a structured course, do the coding somewhere else, and treat this channel as the place you go the moment the statistics stop making sense, because more often than not Ritvik will explain the thing that was blocking you better and faster than the course that introduced it.

[ final ]

The verdict.

A superb free supplement for the mathematical side of machine learning. Keep it in your back pocket and pull up the relevant video whenever a statistical concept in your main course refuses to make sense. As a first and only resource it is too fragmented, but as a gap filler it is hard to beat.