the Maths for Machine Learning specialisation was one of the first that i've ever done on Coursera. and it's still my favourite set of courses i've done on Coursera so far.
Thank you for this recommendation, Giles. I just completed the entire course. tl;dr: I'm glad I did it, it's a great course overall and I would recommend people take it, bu unlike you, the reason I also found the PCA course (i.e. the third and last part) the weakest isn't because the teacher wasn't as personable as the previous two. For those interested, read below for my full impressions of that part: I didn't really mind the fact that the third teacher doesn't really connect with the viewer. Yes, they don't smile much and they don't have the same kind of energy as the first two teachers, but as long as the pedagogy is there, it's all good... But that's the issue really: the teacher and the content make too many assumptions about the students, and the pedagogy felt lacking as a result. Even if the course is labeled "intermediate", it's supposed to be the logical progression of courses 1 and 2, and it feels like anything but that. Mathematical concepts, formulas and notation are often glossed over in a sentence, with few visual or concrete examples, such that you're very quickly overwhelmed if your math knowledge isn't already way beyond what was expected in the other two courses. When you're not already really comfortable with (from my point of view) fairly advanced maths, too much abstraction can be very intimidating and a complete motivation killer. You just end up feeling like you're an idiot or like mathematics just aren't for you. The course content is lighter (4 weeks only), yet at the same time, it covers far more complex topics, and it often feels like much less effort was put into it. For example, the teacher often defaults to giving you reading material - some of which is quite heavy on esoteric notation that is barely covered, if at all - instead of making explanatory videos. Another example: the feedback on quiz questions, most of the time, amounts to "Great job!" or "Incorrect", with no explanations. The quizzes themselves often feel completely disconnected from the lessons that precede or follow them. They can be extremely easy, or extremely hard, usually with no in-between. And the programming assignments can be strangely worded or contain mistakes at times, which can throw you off as a student. They certainly threw me off, even though I'm already quite familiar with Python and numpy. In other words, the substance of the assignments wasn't necessarily insurmountable - I was shocked how quickly I completed some of them, in fact -, but the way they're framed and explained made some of them more puzzling than they needed to be sometimes. All those aspects are big departures from the teaching style of the previous courses. Lastly, though this was also an issue with the other courses, the teachers and staff barely ever reply to questions on the forum. Not too much of an issue when the course itself is great, but for this third course, it didn't help matters. I realize I sound very critical, but again, this isn't to say people shouldn't take the course. They should take it, including the third part, if they have any interest in machine/deep learning and they need to brush up on maths. It's easily worth the €40/month - by the way, you can go through the entire curriculum in a month if you can devote 4-6 hours to it almost every day like I inexplicably managed to do. I just wanted to push back a little against what you said in the video, where you seemed to imply that people were unfairly complaining about some of the assignments. From someone who isn't already very comfortable with maths and abstractions but was diligently making their way through all the content regardless, let me emphasize: it's not just that the third course will challenge you and you should just deal with it; it's that it could stand to be improved, or at least give more pointers to helpful external resources (not just a link to a Wikipedia article). Being challenging or intermediate level isn't an excuse for an unbalanced difficulty curve or hasty coverage of complex topics. I will say this though: that last quiz before the final assignment forced me to do a deep dive on partial derivatives (especially the chain rule) and gave my poor little brain quite a workout. Genuinely thankful for that. And again, the previous two courses alone are worth enrolling for. Looking forward to more course/book recommendations! :)
I am still in the second course, and you wouldn't believe me when If I told you that I already finished my Deep Learning specialization, but chose to study this specialization to gain more depth in the related math, and still I am impressed so far by the first 2 courses.
I am was not strong, when it came to maths at school but in in my "post-high school" education, I worked really hard and exceeded expectations. So much so, that my parents asked if I had cheated or paid someone to write my exams, lol...but this gave me the belief to overcome these fears I had, related to these "high school" shortcomings and changed my perception that DS, ML and AI were realms for other people, not me..... I have become really interested in these fields, ML and AI in particular. I am a long way from reaching my goals but your channel and recommendation are useful, awesome, insightful and inspiring. PS: His new courses starts today and I just watched this vid! 5th July 2021 - Thank You so much.
Actually I completed the first course of this specialization and having completed this far Idk why are people commenting that the assignments are hard. In fact those were totally based on lectures. After all this is the price you need to pay to learn something new and earn a certificate from a renowned institution. Otherwise just fast forwarding the lectures and solving just the petty assignment doesn't add any worth to your learning. Rather it is the best course for anyone who wants to equip themselves with the mathematical knowledge needed for machine learning.
I agree with you 100%, but I do also agree with some of the comments too As in, the course makes it seem like that the course ONLY needs begginer python skills, but you need more than beginner in actuality which makes it difficult then. While one should be able to learn independently, when the course makes it seem like one only needs beginner python skills, it is a bit unfair in that respect. In veracity, its more like, you can start with beginner python, but you will need at least intermediate level. For reference, I finished it too on the spectrum that is that I finished it without much difficulty at all, except for that last assignment. Edit There's also another side to it. The course makes it seem like you don't need much besides high school math, even excluding calculus. And that is true, however, that doesn't mean it will be easy like you stated. This I think can be explained is that people may not be used to this kind of difficulty.
I am wondering why did they not include any course on probability and statistics in this specialization. I took these courses during my bachelor's in engineering but looking for a resource to revise the subjects. Is there any other specialization that includes a course on probability and statistics that you recommend?
Giles, I was sceptical about this video as I don’t remember you ever anything slightly negative about a book or course you’ve reviewed. So I’m glad you pointed out that the third course is the weakest. However, you were still far too generous. Marc deisenroth has exceptionally poor pedagogical skills and as someone who has used, understood and applied PCA in my work for several years, I felt desperately sorry for anyone who came across this as their first exposure. The third course in this specialisation is simply not worth a student’s time.
I just posted my in-depth impressions of the course (the PCA part in particular) and then saw your comment. Well, I'm this person you feel sorry for haha. I don't "hate" the teacher per se - his relative awkwardness is relatable, honestly -, but the way he just throws notation and concepts at you while barely expanding upon or illustrating them is quite frustrating. The reading material he gives - which is taken from the textbook he wrote - is even worse in that respect. He has this way of writing about maths that sounds like it's to the point for anyone who's already a math wizard, but for people like me, it's just cryptic and discouraging. Why the course is the shortest of the 3 parts when it should easily be the longest and most step-by-step, I don't understand. But, as I said in my own impressions, at least I got to drill partial derivatives, so that's nice at least.
Yes there are some challanging labs and programming assignment....but after clear understanding of concept I found easy to solve and the tip is if you exhausted dont do it.......do it with fresh mind.
I enrolled in this course when i could free because of my school account but i now i cant get a certificate without upgrading my account. So the question is if i add the course in my linkedin without the certificate when i completed the course does it matter that i don't have a certificate?
It depends on how you look at it. Assuming, you have not finished the course. If you put it on your LinkedIn and state it in a way that precieves it to be competed in any close or explicit way, then that's just lying. Because by definition, you did not compete the course and the rest it self obvious. If you finished it but can't get the certificate, then with the same above logic, you finished the course, however, can't verify it. However, I don't think an employer will care if the certificate is on or not, especially when you should be examplifying what you learned through projects and real life stuff.
Your justification for why the assignments weren't "too hard" was all the convincing I needed that this class is bad. If I have to spend extra time digging in more depth online then there is no point to take that course in the first place. I'll just save some time and go straight to the research online step. Why take a course where they don't teach you what you need to know?
In the specialisation FAQ, you'll see: Q: "What background knowledge is necessary?" A: "High school maths knowledge is required. Basic knowledge of Python can come in handy, but is not necessary for courses 1 and 2. For course 3 (intermediate difficulty) you will need basic Python and numpy knowledge yo get through the assignments." If you have these prerequisites, you'll find you barely need to do any extra digging online. But do note that "highschool" refers to the British A-Levels, or the American AP exams. The maths in the American SAT is not sufficient. They assume you are ready to enter university, and this is just a brief glimpse into first-year at university. For the Linear Algebra course, they assume you are already comfortable with matrix operations (matrix multiplication, determinants, transpose, etc.). For the Multivariate Calculus course, they assume you're already comfortable with performing single-variable differentiation. They reintroduce these concepts that are required, but quickly gloss over it as it would be more of a refresher instead of a learning point. They quickly move on to more advanced concepts and how to use these concepts in basic applications for Machine Learning. The 3rd course, PCA, builds on these two courses and requires basic Python & numpy knowledge. In particular, being able to work with vectors and matrices in numpy. The course description also says this course is significantly more abstract, so it will require strong mathematical intuition and an understanding of some statistics (mean, variance, and covariance). It's a very good class, but it is a "beginner" class from the viewpoint of a university, not an average person.
What's your thought on Khan Academy? I haven't done math since I was in high school(2009) and I'm revising it from the ground up. I started with pre-algebra, on Algebra 1 now. Why am I doing this? I'm interested in Data science.
Maybe you are doing it so that you can secure a future? It’s important nowadays to master mathematical tools and I wish you good luck. And it’s smart indeed to start refreshing your memory doing high school algebra before diving into college undergraduate courses. It will certainly take time but don’t give up. In only one year, you can get some useful mathematical superpowers. You only get better by practicing. :)
@@duraace6511 Thanks so much for your comment. I've now covered algebra 1, 2 and high school geometry and I'm continuously revising each topic, not so much algebra 1 as you are constantly applying those principles anyway. Pre calculus next. :)
I'm currently planning on taking the IBM AI enginerring course and am considering taking the Imperial College course in order to get my maths up to speed. Maths is far, far, far from my strong subject; I somehow managed to get an o'level and then avoided it until doing an Open University course on maths for science, in which I learnt more than I had in my school years. So... Am I likely to struggle with this course ? Is it a good foundation for ongoing ML & AI study ? Any snesible replies will be welcome. Thanks.
Honestly the Princeton algorithms part 1 and 2 courses are really good and so is the Stanford courses one. MIT OCW intro to algorithms is also really good.
Definitely go for it! Andrew is one the few exceptional teachers who knows what he's teaching. You can tell from the amount of enrollments that he's a very good instructor. Totally worth it :)
I agree with the point of that he's a great teacher and that the course is good and all of that, however, it should be noted that I think the course is better looked at as a way to gain a good foundation of the topic. Deep learning has branches such as computer vision, natural language, etc. He's covers them, however, obviously can't go all the way in depth. After the specialization, you should have a foundation of each branch, at leadt the ones covered, but be able to stem off to which to ever one you want. It should also be noted that I'm not saying that the course is covering the intuitions of each branch only, but I am also saying that it will be better suited for a foundation rather than making you an expert in each branch if that makes sense but at the same time, not a full out noob.
Hey Giles. Thank you for the review. Is this course better than the free Mathematics for Machine learning book that you had suggested earlier on your channel? Is it worth to spend money on this specialization when that free book exists? I am asking since you had given a positive review about that book too.
They actually use in the course for the final projects the Jupyter Notebooks that come with that book (I am not sure if the authors of both things are the same ones). But the book covers the three courses from the specialization, maybe not that deep, and a few more topics.
@@anishray7401 It's been a while since I completed the specialization but I think so. Still, if you start it and you enjoy I think you should complete and having the book as complementary material or a guide for further readings
@@mateogomez8685 thanks again. I was actually thinking of using only book and not the course. My original question was based on this idea itself. If I use only the book would I be able to learn all the required concepts thoroughly? Because if the Book is sufficient I wouldn't want to make an investment on the course since there are other courses that I need to do as well. If mathematics for machine learning book is good I might be able to use it as a replacement for the course. Do you think I can do that?
@@anishray7401 Then, I think the books is okay. You can always dive deeper at some specific topic at Khan Academy, ritvikmath or StatQuest YT channels, for example 🤓
I finished the Deep Learning specialization and would like to take another course/specialization and use my Coursera for students credit on another course. Which of these two is better to put on a resume when applying for a DS/ML/DL job? -Mathematics for Machine Learning (Imperial College London - this one) -Modern Application Development with Python on AWS (Amazon Web Services) I took part in a hackathon, where we used AWS, so I guess some cloud-based experience would be useful in an industry setting, that's why I am considering that specialization.
Having completed the course myself i would say its quite demanding in theoretical level (mathematics) and practical level (programming)...It's not a course for a beginner....
it's 'beginner' from a viewpoint of a university, not the average person. the prerequisite mathematics is only at a high school / college (pre-university) level: some matrix operations and differentiation. it is certainly doable for students who are ready to enter university for data science, and just want to have a brief glimpse into what it's like. in that sense, it's a 'beginner' course.
Can we audit this course for free without the certificate? Will all of the content be available if we audit it? Will the assignment problem solutions be available if we just audit the course?
You can audit these courses. When you audit a course you'll be able to see most of the course materials for free, but you won't be able to submit certain assignments or get grades for your work.
How "bad" are you? I'll just assume in a way that is general. Note, don't just learn how, learn why and and learn even more of the why. For instance, don't just know how to solve for f(x) in f(x) = 5x+7.learn why. Also, learn what it represents, how it correlates to other things, and basically everything else till your brain says f u. Brush up on high school maths on khan, like Algebra. After, you should learn the subjects of linear Algebra and calculus on places like MIT ocw. Where calculus should be single varient, multivarient, and mabey more if you want such as learning about differential equation. But basically, for now, single varient and multi. Or Calc 1 to 3. Where 3 is multivarient. Before calculus, limits will be used a lot in calculus, so make sure you understand them. You can also use helper resources like 3blue1browns RU-vid channel. Desmos graphing calculator for graphing functions too. Also, don't read this like a gospel, just use it as pointers, besides the part where I told you to think of more than just the how. Khan academy should be used for lower level maths primarily, however, more of a secondary for undergraduate and above. Btw, depentednt on how much you want to know about mathematics, you obviously can't stop at calculus 3. But, should be at a point of where you can decide what to learn. As calculus 3 is the last stated material here, it may also be noted that the range of material stated ranges from high school to undergraduate. Edit, I forgot things like statistics and probability and other stuff. For this stuff that I forgot about, go to ocw and you can probulary take it anytime once you start the undergraduate stuff. You can also search up things like course requirements for math bachelors to find more courses.
thanks for the review, but I thought they should have added the Statistics and Probability course as well. I am designing syllabus for Mathematics for Machine Learning for my master's students. I think this is the course which I will consider.
You don't pay for that specific course directly, but you sign up for Coursera Plus (which gives you access to the majority of paid Coursera courses, including this one). The monthly fee is about €40/month, and you can cancel whenever you want. The course took me 1 month to complete, but I had a lot of time on my hands: spent roughly 5 hours a day on it almost every day. 2 months (so, €80) might be a more realistic time frame if you don't have a ton of spare time to devote to it but still try to work at it a bit every day.
it works for me and the link itself appears correct, if its still not working, just go to Coursera and search Mathematics for Machine Learning Specialization
it does very well. I am doing Andrew Ng's machine learning course now after this specialization and it feels like a breeze. But if u are going for this specialization be ready to spend a lot of time outside of coursera on youtube and other places, i found the channel 3brown1blue very helpful compliment to the courses.
@@badreldin2 i was a bit worried that it doesnt cover probability and statistics and that the course on linear algebra is shorter than most university courses
@@QasimKhan-nd8og you could do the "Advanced Statistics for Data Science" specialisation by Johns Hopkins University, but it's quite difficult and is more of a pure statistics course instead of applying it in data science.