I have already completed two courses in MLS, will that make sense if I take Math for ML now or after MLS?
I think you can easily do Math for ML after MLS. Andrew briefly explains the meaning of the math concepts used in MLS and then mentions you don’t need to understand them to go forward with understanding the ML concepts.
In the discussion with Chris Manning, at the end of MLS Module 2, Chris emphasizes on the importance of understanding the math if you want to take the subject seriously as a profession, but that it is not needed to get started with the domain.
Andrew then provides an analogy, in that he is not good at quantum physics, but he can still use a computer, although not understanding the physical challenges of a working transistor, since he wouldn’t need to fix the transistor if it brakes.
Not sure how important it is for you to understand the low level explanations in detail before going forward with these concepts, but I don’t think you need them to understand what’s being taught in MLS.
It depends on your goals. You can certainly take the M4ML course later.
For me I think there’s not a lot to be gained from the M4ML course if you’ve already passed other specializations. It’s intended to provide the math background for comprehension of ML methods. If you already completed some other ML courses, maybe you already have the knowledge.
Hello @Kesavan_Ramalingam , personally, I have taken many specializations including MLS but I still recommend Math for ML (E.g. the relationship between MSE and MLE is very well explained). Some notions of linear algebra are also very well developed. Luis Serrano is really a master of the subject.
Assumptions are also very important (e.g. conditions for applying linear regression) and are often overlooked in ML courses. Sampling from a distribution is also very well explained. So I’d say go for it, take the course, you won’t regret it!
Note that the teaching style in the M4ML lectures (which is very good) is somewhat inconsistent with style of the programming assignments, which some students find very confusing.
The programming assignments also use different conventions and notations than the other DLAI courses, so be aware of that going in.
Hello @Kesavan_Ramalingam,
I am trying to list out some facts with my comments but the decision is yours. Btw, I assume we are speaking about the Maths for ML and DS specialization. For its name, you see that it isn’t just about ML.

MLS and M4ML&DS are not prerequisite to each other, but they exist in both to some extent. MLS walks through major ML concepts of maths significance in a more storytellinglike way but requiring us to accept the maths outcome, so we won’t be asked to derive any maths formulae or asked to derive meaning from a maths formula  the formulae and meaning are given.

Usually learners are not satisfied with just being given as facts, but wish to go deeper for many reasons  I am sure you have your own if you are one of them. If you accept all the given facts and find them enough to deal with your problems, then you might consider the M4ML&DS when you feel any need to go deeper (e.g. when you teach or explain ML / when you do ML research / when you read ML research, …)

The good chance of learning the content of M4ML&DS is that, you can appreciate better the application of math in ML, for example:

Upon course 1, you will get to see those Dense layer alone (without activation) as simply some kind of linear transformation. That explains why we need nonlinear activation to introduce nonlinearity. That explains why all consecutive Dense layers that are not separated by any nonlinear activation is effectively collapased into one Dense layer. That gives you some idea of vector space, and learning general idea of vector helps you “swallow” a lot of important concepts such as “embedding” in the future.

After course 2, you will find out why gradient descent works at all, why we need to take the first derivative at all. You will find out how people calculate derivative if you need to do it yourself too. You will also come across cases with the second derivative being involved which is never introduced in the MLS and why they may help.

Course 3 will show you something you almost won’t need for MLS, especially the last week of it about hypothesis testing which I am pretty glad that they have included it. Machine learning models are statistical models, via gradient descent, looking for a set of trainable parameter values that makes its predictions mimic the observations (the true label). Probability is a tool for statistics work, without knowing much about it, though, as long as you accept the outcome, you can use any developed tool to carry out your work, but you may also find bottleneck when you want to do statistical analysis on your dataset not just for building a model, because those analysis requires your correct understanding of the statistical model assumption you are making, and how to do the steps correctly, and more.

What you can get from the courses are not limited to what I have said. I am just glancing through the TOC of the courses to give some quick and possible takeaways. Even if you can’t get everything after finishing the M4ML&DS, the courses should open you the door to explore more on your own.


Your learning style. At school/college, we always take multiple subjects at the same time  like we take maths and physics in the same year knowing that they help each other. At work, time management is an additional concern; usefulness to your immediate target is a concern of your precious time; effectiveness is another concern meaning whether you will have time to work on some exercises or projects or revisiting MLS for second time, or any thinking process that lets the knowledge from MLS and M4ML&DS to do chemical reactions in your brain to make the knowledge finally yours (or sink in). It is not unusual to hear people feel not diving deep enough after finishing the courses, which can be due to different reasons for different people but if you think my general suggestions for chemical reaction are not your ways out, you might want to be prepared on how to make the most of your time completing another speciailization.
@Kesavan_Ramalingam, I wish you good luck, and please consider my reply as something for you to think about. I don’t know you well enough to give any specific suggestion, but most importantly, you must be the best person for making that decision.
Cheers,
Raymond