How to learn math for data science (the minimize effort maximize outcome way)

yes you do need math statistic linear algebra and calculus so if you're panicking because you've never been good at math and have hopes of becoming a data scientist let me tell you a story first my story math has always been one of my

weakest subjects math along with english were my worst subjects in high school and here's my college transcript notice that i didn't do calculus 2 in the next semester of my first year like i was supposed to but instead did it in the summer that's because i was doing so poorly that i dropped the class i've always been good at biological sciences but noticed physics and physical chemistry aka the more math heavy sciences they were not my strong suit i also never did multi-variable calculus or linear algebra now fast forward to my master's degree and i go through my experience of failing my first ever test in well you guessed it math discrete math and probability to be specific but you know something interesting happened in grad school and i eventually became a ta for the very class that i failed my first ever testing no i didn't suddenly get hit by a bolt of lightning or get beat up and woke up as a math genius like this person but i started to become more confident in my abilities i don't think i would ever be a genius math person nor would i ever come easily to me but i think my biggest issue is that every time i saw math i would literally internally panic so much that i couldn't even focus on doing the problem and i just kind of assumed that i won't get it and when people try to explain it to me i'm just like i don't get it but what i realized is that if i can just actually get my [ __ ] together for like two minutes and read through the problem and listen to what people are saying slowly it started to make sense remember the

most important trait of a data scientist is to have a growth mindset the belief and the will to improve your abilities and skills now let me put in context how to approach learning to math if you literally freak out when you see math like i did what you need to do is understand things at a high level first and then do just enough to understand the logic behind a concept and if it's an algorithm what the assumptions are and where it performs well and where it doesn't i've given a brief example for a k-nearest neighbor in my how to learn data science in 2021 video which i highly recommend you check out because it gives you my full recommended approach so assuming you watch that video and are now convinced that you gotta do projects and you decide to go do a project let me give you a full example of how to learn the math say you come across a dataset that has lots of information about cats

let's say you want to predict cat tail length by their body length because clearly that is what we're all interested in let's say you start off with the simplest linear regression is the same as the equation that we all learn in middle school math which is y is equal to b one x plus b where x is the cat body length and y is what we're trying to predict which is the tail length and b one are coefficients that determine the slope and intercept of the model so how do we solve it there's a method called a least square solution and what is that well that is when your goal is to have a data set like this one and you want to fit a line through it such that it minimizes the distance between each data point and this line you're drawing which has the equation y is equal to b one x plus b so now you can dig a little bit deeper and ask how do we iterate through all these possibilities for b1 and b to find the best solution and that is how you get reacquainted with calculus i'll link josh starmer's amazing stat quest playlist about this because he explains it far better than i could ever do it and then you're like can we do even better to predict tail length and the answer is yes we can in our awesome data set we have much more information like

hair length for example and that is called multivariable linear regression because you have other variables in addition to just body length to predict detail length and how do we solve that well it's pretty similar to what we were solving earlier but instead of fitting a line through it we're now going to be fitting a plane because each additional variable is another axis on our graph this is getting exciting what if we had even more variables like gender and cap rate would that help us predict hailing even better but wait gender and cap breed are not continuous variables they are categorical variables interesting so how do we deal with that you see this is exactly how you can acquire these math skills in the context of solving real problems i'm personally so much more motivated than in math class when they just told me to solve random problems for no apparent reason another key reason why project guided learning is amazing is because you have a guide to what it is that you need to learn and you're not just blindly learning a bunch of information that may or may not be even useful you also have the choice to dig deeper into concepts if you want to and i need to emphasize this you can dig deeper into concepts only if you choose to do so because there's always going to be more math and more rabbit holes to go down you are in control of what you want to learn and i always say just to learn enough so you feel comfortable understanding a model and using it which for me for linear regression is understanding that you're trying a line to minimize the distance from each data point to the line and you can also have multiple variables where you add in additional dimensions for each additional variable also the assumptions of a linear regression model are homogeneity of variance independence of observations normality and linearity and finally it's important to know when linear regression is useful and when it does poorly it's another homework question for you guys so comment the answer below so maybe you're done with this project and next time around in another project you're again using linear regression maybe then you can dig a little bit deeper into the math or maybe at some point you realize that there are even better models and you go and you find out how they work as well i think a huge misconception is that people think math is equal to doing math problems and manipulating equations and that is certainly part of it but unless you want to go and fiddle with formulas all day what's important is to understand the logic of the math nobody's going to go and ask you to solve calculus questions but what you should know is understand how gradient descent works for example which is calculus and that my friends is my technique for how to build up your confidence and learn the math you need to become an awesome data scientist i also want to mention that in your day-to-day life these math things that you know are often not even consciously in your mind it's just part of the knowledge that you have in your brain that will help you confidently solve data science questions i hope this was a useful video for you guys and you feel a little bit more confident in your math skills i'll see you guys in the next video