- 5 reviews
- 4 completed
Content: Covered a broad range of topics in modern day machine learning; but most importantly it started from a fairly simple concept - how to iteratively solve a derivative - that anyone having done higher level maths in high school would have been familiar with already. From there on it once you get that then it is just adding on various tweaks and related concepts. You get a focus in this course on the practical part of doing such analysis - that is the problems/pitfalls of each technique, what are its uses/misuses, how to "tune" the parameters of each technique to your liking etc. So all very useful for someone looking to actually implement this stuff, in real life! Instructor: Prof Ng is first and foremost an extremely good lecturer - I have never experienced US-style education but thanks to this course I badly wished I had done my undergrad in the US. Always breaking down difficult/annoying concepts to its basics, going through carefully each equation - he is exactly who you might need to get through such a technical topic.
This course was insightful, in the sense that I have never really considered the "messiness" of real data that was not cleaned up and filtered extensively by the guardians of financial data (Bloomberg and Thomson Reuters Datastream come to mind). It was a fairly broad brush class that covered a whole variety of different methods to obtain data via internet, SQL, pictures etc but probably the best takeaway of the course was the introduction to the concept of "tidy data" as espoused by Hadley Wickham. I feel that that is a standard well worth keeping in mind as we try and navigate massive data sets - it is too easy to mess around so much with data that it makes no sense to anyone else, even a fellow professional/expert!
My first course that actually just focused on programming in R, as opposed to using R for statistical analysis. I have no education whatsoever in computer science/programming so needless to say that some of the concepts/homeworks were quite painful for me...however I felt that it was worth understanding because you come away with such a better idea of R's capabilities and usefulness, which are of course totally transferable to other courses that require R.
So this was the second course I did after having completed Andrew Ng's Machine Learning course and wanted to go down a bit more into the rabbit hole....this is the first course of the "Data Science" path for John Hopkins and I would think of it as just a very gentle introduction to Github - which allows people to work on a project, collaborate with others and have a very precise version timeline on the project, which frankly people doing shared excel or powerpoints could learn a lot from. The course is quite easy but don't stop there - this is just the tip!
This was my second course on machine learning after having taken Prof Ng's Machine Learning - and oh boy, the teaching content and techniques could not have been more different. But this is not a bad thing at all - rather this course takes a much more rigorous theoretical approach towards machine learning. You get to learn a lot more about the motivation behind doing one sort of technique vs. another - and indeed they look much further into the algorithms than on Prof Ng's course. One thing to note is that all lessons (and homeworks) will require R - as opposed to Prof's Ng's course requiring matlab/octave. I believe R is one of the de facto standards for open course statistical analysis and so highly recommend that you take this opportunity to learn R in the course (or maybe take a course focusing on R alongside) rather than see this as an obstacle. Homeworks are a key part of learning for the course - you get quick questions for every video + section quiz for each week's content. Do not underestimate the difficulty of the homeworks, particularly as not all of the questions are well phrased (and this was hotly debated in the forums when I did this course earlier this year). Luckily you have until the end of the course to do all of this in! Despite having said all that about the homeworks I still highly rate this course - knowing the theoretical underpinnings of the most common machine learning techniques taught by world class professors in their field was extremely enlightening for me and their (free) pdf "An Introduction to Statistical Learning" is highly accessible (I have no science/machine learning background and found most of it intuitive enough) and has lots of R examples to follow. And this is all free!