Effective ways to practice?

Course Q&A Math for Machine Learning M4ML Resources
1

Hi, just wanted to first say, I learned a lot from the specialization! That said, I was wondering if anyone can help with suggestions for effectively retaining math topics I’ve learned?

I find that I can learn concepts decently and understand how to use them when practicing examples, but I find I have a trouble retaining this knowledge. Right now, I’ve been practicing problems a few times a week on KhanAcademy starting with Calc BC to try to build up long-term familiarity, but even now I’m just practicing on those things I’ve seen more often in some of the ML classes I’ve taken. Even this takes a while, and doesn’t even encompass all of the problem sets obviously. On top of that, I still need to do the same with Prob & Stats, and also Linear Algebra, which will be more difficult to practice since there aren’t any randomized practice problems for it for individual topics on Khan.

Essentially, I don’t know how to go about practicing all the math I’m likely to encounter in a manageable way, without spending over 50% of my available learning time just on internalizing math. Not to mention, if I spent 2 hours a day just to practice math without using it in some constructive way (that isn’t just a project “for the sake of learning” but in an actual use case), I feel like I’ll go crazy, get sick of it and lose motivation, especially since math is a means to an end (ML) rather than me doing math purely because it’s the thing I enjoy. And while I don’t have anything against learning everything I need in regards to math, I do think it shouldn’t feel like my life should just feel like I’m doing nothing but study math, and currently, I feel like this is the case unless I had a magnet for retaining all the equations and relations and procedures for everything I’ve learned.

I just think there are so many topics, and lots of equations, problems, and workthroughs take so much time to do just one, that it’s difficult to practice more than a handful of topics a day, or even a week. And then I’ll get to understanding it, go to practice some other things, and maybe in a few days or the following week, come back to the same topic and then find out I have to “relearn” it, and so on and so forth. All while also trying to learn new material and adding to the carousel of material to internalize. How can I get practice and reiteration on everything I need, without feeling overwhelmed that I’m just relearning the same things over and over again, but also in a manner that I’m able to review everything I need to know, without spending all my available learning time purely reviewing math instead of other ML topics? It’s hard for me to retain something unless I’m constantly exposed to it, but at the same time, during the other courses I’ve been doing in regards to ML, it’s not like those topics are reviewed every lesson or every course, but rather come up on occasion, by which point I’ve already forgotten the concept/equation and I sit there knowing I learned it, but feeling pathetic because I learned it just to now see it, knowing I spent the time to learn it, but not remembering how they got to this formula or that.

1 Like
2

Frankly, I think very little of the M4ML course applies to actual Machine Learning work. You’re never going to have to use Gaussian Elimination, analyze the probability of loaded dice, or any of that.

There are some useful topics, but many others are traditional “math for fun” concepts.

I recommend you start taking Machine Learning courses, and whatever math you need in those courses will be immediately apparent. You can practice the math by doing the course work, and I think that will help solidify your knowledge.

2 Likes
3

Hello @aesyrin,

If you want to solve problems with what you have learned, then just answer questions in this forum. To be able to do a good job at that, you may need to understand the concept from different angles, since learners see things in their own ways.

Of course, those angles don’t come out of nowhere, but through thinking and sometimes trying to re-understanding the concepts. There is a chance that you will find angles that are related to some ML topics of interests.

If you are bored with pure maths, then when answering questions, try to be creative and bring in something both interesting (to you, like ML examples) and relevant to the question.

Lastly, it does not matter whether the question you would like to give it a try was answered or not. You can always share your answers and if you want to draw more attention, please feel free to open a new thread, and along with your answer, some quick description of the question you want to tackle and optionally with a link to the original thread.

I look forward to your participation!

Cheers,
Raymond

2 Likes
4

Thank you both for your replies and suggestion. @rmwkwok , that’s a good point about re-understanding something to be able to explain it to someone else. I’ll definitely need to set aside some time to do that here and there, and on top of that is has some benefit to others besides just getting practice out of it.

1 Like
5

Sure! Do give it some trys, @aesyrin! Maybe it won’t start cool, but that does not matter, because, over time, what you learn in this process is all yours and will become some strong foundation for you to move on! If you let us, we can be with you in this not-necessarily-cool initial stage :wink:

In fact, the less cool it is initially, the more room and value it represents. :wink:

Cheers,
Raymond

1 Like