The Math of Machine Learning: Do I really need it ??

As I covered in my previous post, math plays a very crucial role in one's path to mastering machine learning, but what if you don't want to master ML ? I'm no purist, and I totally understand that there are people out there who're not willing to dedicate their lives to this, they're just enthusiasts who want to get their hands dirty. Well the fun and exploration is what keeps people (who're not in the field directly) interested in coding.

So what if you're an enthusiast or someone who's into business, wanting to learn about the new buzz words in town. Do you really need to learn Linear Algebra, Statistics and Probability to understand ML ? Not really, you just need to have a knack for grasping concepts. You see, most of ML breaks down into simple logical blocks where your data is processed in intuitive ways.

Here's an example,

Take Linear Regression (something you've heard enough of already I guess), according to Wikipedia, it's defined as:

In statistics, linear regression is a linear approach for modelling the relationship between a scalar dependent variable y and one or more explanatory variables denoted X. The case of one explanatory variable is called simple linear regression.

 

Oh well, it has symbols like sigma which most of us have seen before, but still unwrapping it feels so complex. Now let me offer some help.

Hmm, so it looks like you've your data and all you're doing is having that line in red line up perfectly such that it goes through all of them right ?

Here's another...

Take Gradient Descent, according to Wikipedia, it's defined as :

Gradient descent is a first-order iterative optimization algorithm for finding the minimum of a function. To find a local minimum of a function using gradient descent, one takes steps proportional to the negative of the gradient (or of the approximate gradient) of the function at the current point.

Now unless you're some math wizard, chances are you didn't quite get this. I mean half the symbols would've looked new to most people. Now what if I told you, this is a very simple concept when you look at it graphically. Here let me give you some visuals...

Now when you look at this and compare what you read, it's easier right ? All you're doing is finding the lowest possible point in the plot. Kinda like rolling a ball down a slope right ? See, how many PhDs did that take ??

Ok, now I've given you a few examples, well that helped you understand those topics right ? Right ? Wait, you still don't know why and where they're exactly used right ? See this is a good starting point, most people don't even make it this far, seeing such equations and statements they give up, but atleast this way you get to understand what really is happening, instead of rigorously (regression pun intended 😜) grinding through scholarly articles and such to understand such topics.

So the next time your ML friend spouts how he hasn't been replying to your texts as he was busy tweaking his hyper parameters, you know why (go research 😜), you know how to.

This is an approach I'd recommend to people who're just checking out ML, for the hardcore purists, dive in the way you find comfort in, as you'd want to build full understanding of everything ML.