The Math of Machine Learning: Noob's Guide



So after seeing all the hype around AI, you've decided to get started with AI yourself. The idea of making yourself a personal assistant and becoming Tony Stark sounds cool right ? Well it is, trust me, I've been there and done that already. While it's fun, to whip up some chatbot api like the Facebook messenger one, upload a bunch of sick scifi lines for your bot to say, after a few weeks the amazement ends. And the bot will probably be forgotten like most other fads.

Or maybe you were one of the cool kids who picked up a library like scikit-learn or tensorflow for your project. Well the getting started tutorials have probably been exhausted at this point and you're left wanting more...

So what do we do then ? Most tutorials sound complex once they get into the math. What's gradient descent again ? What's this sigmoid function ? What's all this Latin doing in my English !?!




So for those of us who can barely integrate functions, here's a guide to get started with the math for machine learning:

Linear Algebra: Concepts like regression require you to have an understanding of linear algebra. You need not master every concept, just have an idea of what goes on and why.

Statistics and Probability: You must've heard the word 'predict' every few seconds in any ML video, so probability is obviously a requirement. It's advised that you make a cheat sheet of all the formulas and concepts you come across for reference. Remember you're not giving an exam (or are you ?!) so it's fine. Now understanding graphs is a must for you to plot, and work on your data so the drill's the same as with probability.

Should I just get back to college then ??

Not really, sure some of these must've been covered in your college days, but that doesn't mean you've to get back to college. I'd suggest checking out Khan Academy, Udemy, there's quite a lot of resources out there for enthusiasts like us. And the quality (not referring to the video resolution) is just out of this world. If you still want to take classes from a real person then that's up to you.

Is there a way without math ? 🙄

Well, you'll need math at some point or another, otherwise it's pretty difficult to scale and grow here. So I'd highly recommend giving Linear Algebra atleast a try. As it's fairly simple. You could always partner up with someone who's good at what you don't know and work together. I'll cover this in detail in a separate post, so stay tuned.



Comments