Exploring the mathematics behind machine learning magic
The math behind machine learning can (and should) be explained in a way that's easily digestible for non-expert practitioners.
Welcome to The Math Behind the Magic.
TL;DR - If you’re intermediate/late beginner at machine learning and want to start taking the first steps into understanding the math behind it, follow along as I learn this stuff too.
Why this Blog
I’m currently taking a mathematics of machine learning course in graduate school, similar to courses taught by Ioannis Mitliagkas, at Princeton and MIT. My background is more on the applied side. I built data science models in industry for a few years and took all of the typical intro machine learning curricula. This course is the first time I’m asking myself why these algorithms work in addition to just how and when they should be applied. I wouldn’t say I’m ignorant to mathematics, but I’m not as comfortable with its theoretical explanations. Up until now, I’ve learned just the right amount of math that I needed to.
Machine learning and the concepts behind it is not new. What’s new is the infrastructure (GPUs, cloud computing, etc) that enables folks like me — who primarily care about getting a model to production — to get started with machine learning right away. The downside of this is that it takes about 5 Google searches per word for me to understand a theoretical machine learning paper. But I recognize that in order to succeed in this field in the long term, you will eventually need to understand the math behind the magic. That’s why I took this course and why this blog exists.
What to Expect
Expect a slow, step-by-step explanation of math-heavy machine learning papers with lots of pictures and graphs. In my opinion, the ability to visualize math beyond the written symbols is the only criteria for truly understanding it (and that is not easy). After all, humans are visual creatures.
Expect a perceptive of someone who’s not an expert at this. I hope this perspective is useful, since my perception from my studying thus far is that mathematical explanatory papers are written for mathematicians by mathematicians. I don’t think this is a bad thing, but for a field where the entry level skill set does not require math, I believe there should also be a space for those who are not experts but would also like to have an understanding beyond a novice level. Expect me attempting to curate such a space on this blog.
Expect me to not keep up with this blog when the semester is over, i.e, around May 2021. But in the meantime, I hope we can learn a lot together!
Subscribe to Tag Along!
And tell your friends!