How to Learn Machine Learning: Top Down or Bottom Up?

by Justin Skycak (@justinskycak) on August 05, 2024

It can be helpful to take a top-down approach in planning out your overarching learning goals, but the learning itself has to occur bottom-up.

Want to get notified about new posts? Join the mailing list and follow on X/Twitter.

Many people try to study machine learning by diving straight into the deep end.

Typically, what happens is this starts out very exciting…

but if it turns out you’re missing lots of foundational mathematical knowledge, that excitement quickly transitions to confusion, flailing, and ultimately quitting.

Now, I’ll admit that it can be helpful to take a top-down approach in planning out your overarching learning goals.

You want to learn machine learning?

Then, looking backwards, you need to learn most of single-variable calculus, linear algebra, probability & statistics, and a bit of multivariable calculus.

And looking backwards even further, you also need to learn high school algebra if you don’t know it already.

But the learning itself has to occur bottom-up.

Math is a skill hierarchy, and if you cannot execute a lower-level skill consistently and accurately, you will not be able to build more advanced skills on top of it.

Trying to master calculus without having mastered algebra is like trying to practice a backflip despite barely being able to jump. All that will happen is you’ll land flat on your face over and over again until you get sick of it and give up.

Now, I’m not saying you shouldn’t occasionally watch machine learning videos or have some fun playing around with an off-the-shelf model if that helps you build motivation to train your math foundations.

But if you don’t also work on developing your foundations, then you will never develop a high degree of expertise in machine learning.

∗ ∗ ∗

Want to get notified about new posts? Join the mailing list and follow on X/Twitter.