- 9 reviews
- 7 completed
Professor Lewin's 8.01 is probably the course most responsible for why we have MOOCs today. I remember reading about this course in the NY Times years ago and it was one of the first offerings from MIT's OCW more than a decade back. It's about the clearest and most exciting introduction to mechanics one will ever find. Lewin has very very few peers when it comes to instructional talent: the only others I have seen that are as good are Terry Tao, Richard Feynman, Harold Abelson, and Ronen Plesser. Lewin tells you he will make you love physics, and he really means it. Prof Lewin's lectures aren't filled with tedious derivations that are best left to the textbook. Instead, they're mostly tests of whether physics really works, or whether it's just some BS that looks nice when you calculate the answer on your problem set. The first thing Lewin does is teach you how to deal with uncertainties so that you can meaningfully analyze an experiment. The answer to how far the ball flies never agrees exactly with the one your calculator spit out, but it's such an awesome thing to see how you can factor in your uncertainties due to your meter stick, due to your ability to stop and start the timer, due to how close you could set the initial angle, and so on. Once you learn this, the true aim of the course shines through. Lewin treats the course almost as if you should be a skeptic, and then he goes and shows you these laws actually work in the real world to very well-defined limits of uncertainty. His aim is to show you that physics works, and that it's not just doing mathematics. Lewin loves to throw very difficult challenges out at you at the end of lectures to make you think of difficult cases of problems discussed in class. He always says it's good for a budding physicist to have sleepless nights thinking about these problems. But the crazy thing is, he gives you the knowledge you need to solve these. Where this course really shines most is in the sections on angular momentum and torque. These are very difficult concepts treated with great care, and lots of physical demonstrations. Furthermore, Lewin can throw tons of interesting rotational examples out at you involving astrophysics (his field). The spinning of an ice skater is nice, but the collapse of a neutron star is a way cooler angular momentum problem! When he illustrates precession with a bicycle wheel and shows the consequences of the conservation of angular momentum it is so unexpected, and it would be such a difficult thing to accept if you only read it in the textbook. To me, we live in such a great age to be able to take a real MIT physics course from one of the best teachers to ever live. Anyone who likes physics will just fall in love with this spectacular course that always emphasizes the physics, and never focuses on just computing the integral. This is the course that blazed the trail for everything we take now, and it's still probably the greatest ever made. Not to be missed at any cost.
This is a pretty advanced introduction to multivariable calculus, taught from the mathematician's point of view as opposed to the engineer's. With a few engineering multivariable courses already available online (for instance, Denis Aurox's brilliant 18.02 course at MIT OCW), it's really nice to see one that treats the derivative as a linear map, matrices as representations of linear transforms, kth derivatives of real valued maps as k-linear forms, and so on. It's not a complete introduction to linear algebra by any means, but gives you enough to really understand differentiation while giving some tools that should help towards constructing a theory of integration in Rn in the upcoming follow-up course. There is a lot of material in this course and you are asked (but not required for credit) to prove lots of the theorems, so you can get a very good intro to the subject. Not quite Calculus on Manifolds level since occasionally a critical result will just be quoted (e.g., the implicit function theorem), but it's still a pretty well-developed theoretic treatment of multivariable differential calculus. I really would like to see the implicit function theorem added, as well as a discussion of finite dimension inner product spaces if/when this course is offered again, as I needed both to prove the result on Lagrange multipliers. With that said, some people will be put off by the focus on solving problems as you read the material; there is almost nothing in the way of videos. The course is more like going through a really well-written workbook. I really hope Jim adds some more video in future iterations, because sometimes it really helps to have an expert derive a difficult result or execute a difficult computation in front of you. But the learning activities that form the real meat of the course are outstanding. There are theoretic problems, computational problems, and it's a great way to learn by doing instead of by watching. I fully agree with Jim and Steven that this is much a better format than the typical 1-3 hours of weekly video you usually see on coursera, though I think some video on the hairier parts would still be pretty helpful. You can tell this course is a ton of work for Jim and Steven. It's a bit rough around the edges with bugs in the activities in this first run, but those should be mostly ironed out by the second offering, since they're corrected pretty quickly here in the first. Another big plus to this course is that Jim and Steven are both very active on the board answering questions. You can tell these guys are enthusiastic about the material, and I'm really excited to see where they go from here with their future offerings. Jim and Steven both sound like kids in the candy store when they talk about future courses they'd like to offer. I hope they can do them in this style. Overall, even with some of the omissions I pointed out above, the material is really incredible and the problem solving approach to me really helps to make the difficult topics like multilinear forms fall in place. YMMV, but the material presented in this format taught me a lot.
This is a quite difficult course, especially the final three weeks when you discuss the Lp spaces and their duals, then study distributions and Sobolev spaces, and finally end on PDEs and finite element approximation. I mean it's a real treat to see a course at this level taught for free in a very theoretical style (the so called French touch), and yet you end up touching briefly on a finite element method for solving PDE like it's done in the real world when you don't have simple closed form textbook examples. Really cool material all the way around. There is about an hour a week of lectures, which really aren't enough to understand the whole course. That's not to say they should be skipped. For instance, Prof Cagnol gives a very intuitive explanation of the Lebesgue integral as summing up horizontal slabs under a curve to contrast it to the Riemann integral that sums vertical columns under a curve. It drives me absolutely nuts that you don't see this clear and illuminating diagram in some of the standard first year grad analysis books like Big Rudin or Royden, though Folland thankfully does seem to understand how much insight this provides when integrating simple functions. The lecture notes and the peer graded problems are the real meat of the course, and while the lectures give some great geometric intuition, they're pretty tough without working your way through the notes also. I think the prerequisites are understated. To succeed in the peer review problems a first real analysis course is a necessity. Functional analysis is way too advanced to be learning how to prove your first theorems. Some exposure to Lebesgue integrals is pretty important too. Not saying you need to have worked through Big Rudin or anything, but you do need to understand things like convergence almost everywhere and similar things relating to sets of measure zero. And 4-6 hours a week? No way if you're going for the advanced certificate (>= 80%). Probably not even if going for the 60% certificate, since almost 40% of the course grade is from Peer Review problems, which are all proofs. Overall, I feel like I learned a lot, even though I have a bachelor's in math with a couple of analysis courses under my belt (but no functional analysis before this course). It's a really enjoyable course if you like analysis, and the PDE capstone at the end really ties everything together beautifully. I really hope to see Prof Cagnol and Prof Rozanova-Pierrat give us some more French touch math on coursera, and hopefully they'll do this wonderful course again too. If you don't think this course is beautiful, you have no heart.
Algorithms: Design and Analysis, Part 1 is an interesting course covering some of what Prof Roughgarden calls the greatest hits of computer science. It's focused much more on math and correctness than is Sedgewick's Algorithms series, which deals with concrete implementations. So they are very different courses despite having similar names. Unfortunately, this type of course doesn't translate nearly as well to the MOOC format as the more implementation centered course by Sedgewick. While Tim Roughgarden is a very good teacher and he gives you the same lectures he gives his students in Stanford taking the class (the actual Stanford lectures are available online and not too hard to find in google), he cannot give the same level of assignments. In a course like this you really need to do lots of proofs and derivations, so Prof Roughgarden should really employ peer grading in future iterations of the course to make it more like the real thing. The assignments aren't as good as the lectures. The quizzes are multiple choice questions while the projects require algorithm implementations where you type in a numerical answer that can usually be arrived at even with lousy implementations of the algorithm, and often even with other much more inefficient but straightforward algorithms. I can't hold it against Tim because it's not a reflection on him, but on the limitations of the online format. I hate to sound too negative though, because I really did learn a lot from Tim in the course and the lectures are of very high quality. Overall it was a great experience taking the course. One thing I really thought was awesome was that he gave optional advanced lectures that weren't required to succeed in the class nor to get an excellent overview of the subject, but that filled in details in the more difficult corners of the topics discussed. Overall I'd highly recommend the course, but would also suggest having a textbook like CLRS to work problems out of while you go through it. My personal highlight in the course was Karger's algorithm for finding minimum cuts in graphs, as it's often not taught until graduate level in a course on randomized algorithms.
I dropped the course when for two weeks straight the homework consisted of reading proposed solutions and saying if they were correct or not. Maybe it was due to rounding errors the autograder had with the first week's two assignments. But there is no way I can recommend a course with ten assignments where four of them are just reading some short answers and saying if the arguments are any good.
This had the potential to really be an incredible course. The professor gives excellent lectures and the material is just mindblowing. Unfortunately, Galaxies and Cosmology was not a complete course in Winter 2013, since there were no assessments other than the video quizzes used to ensure the Caltech students watched all the videos before coming to discussion section. If you googled the Caltech course webpage you could get the homework assignments, but they weren't presented on the coursera page itself (at least in the first 5 weeks or so of the course when I was actively taking it). You're on your own for the homework assignments that can be downloaded from the caltech course webpage. Perhaps this is designed to keep Caltech students from being able to easily find answers on the coursera boards. If you're willing to buy the textbooks and put in the time to do the readings, as well as putting in the time to do the homework from the Ay21 caltech page, this is probably a lot closer to a 5-star class than the three I rated it. But the coursera course alone gets 3 stars for incompleteness. Overall, I learned a lot in the first 5 weeks I took the course through doing some problems out of Ryden. I ended up dropping mostly because there were a couple of CS courses I REALLY wanted to put time into that started about the time I dropped. I might try to finish the course this time though (offered again Janaury 2014). Forgot to mention George is pretty active on the boards, so that's a big plus. Also, make sure to have pen and paper in hand watching the lectures. The math isn't overly difficult, but there is a huge amount of material presented quickly (it's Caltech after all), and without following along in the calculations and filling in the details it would be really easy to get lost.
For anyone who loves programming, this is a truly memorable course taught by one of the great heavyweights in the field. It's really something to see that such an important researcher as Sedgewick is also a phenomenal teacher. It's beautiful the way he shows the algorithms operating. Most powerpoint/pdf lectures are just frankly wastes of time to me, but Sedgewick animates his algorithms slowly and without skipping steps on his slides so that you really see how the algorithms work. It can be so hard to look at a piece of code and decipher exactly what the process is that's being executed, but the animations really help to clarify. A lot of these graph and string processing algorithms can look pretty daunting, but Robert has a gift for explaining them in a straightforward way so that you can remember and understand them. I also love that Sedgewick is very concerned with real world performance vs simple asymptotic results like you see in books such as CLRS. Aside from the quality of presentation, the quality of assignments is outstanding. The seam carver program alone is worth taking the class for (seam carving is a way to find the least important parts of an image so that they can be cut out without losing the main action in the photo when resizing; the results you'll get with the program you write are incredible!). And as previously mentioned, the Burrows- Wheeler program is great too. The autograder is spectacular, since it times your programs as well as tracking their memory usage. The autograder was responsible for building up a great community of students in the forums; we would post our runtimes and memory usage as a challenge, and two hours later someone would post better ones. And then we'd all tweak our algorithms to try to beat that runtime, and so on. I mean it's highly addictive to see how far you could push things and how much you could improve the textbook implementations once you got the program correct first. The one downside to the autograder is that it is so good and detailed that it discourages testing, which just in the end makes your program take longer to write. Even though I think this is a major trade-off, it's probably still worth it to have that competitive (but in an enjoyable way) atmosphere in the forums that would never be there if submissions were limited. One other item of note: Professor Kevin Wayne was EXTREMELY active on the forums. I can't count how many times I'd have a question about something and Professor Wayne would weigh in within a couple of hours to answer me. I don't know how he has the energy, but we'd be fools to not take advantage!
Ronen's Intro to Astronomy is probably the best MOOC I have ever taken. I was skeptical when I signed up because it's algebra based, but it definitely wasn't the kiddie intro I was initially expecting. Instead it's an unbelievably fascinating look into our universe just filled with mind-blowing discoveries, and current ones. Current as in this year or this month. Though the mathematical level is pretty low, there is a ton of information in this course. And it's not rote memorization (if it was I would have hated this class and rated it an F-); Ronen gives very clear and convincing explanations, though he talks fast so you might need to pause every once in a while to catch up! Just so much cool stuff in this course, and it really gets amazing after a few weeks in when you start learning about the formation of the solar system, then the main sequence, but the supernova, special relativity, and black hole sections are the real sweet spot of the course. Ronen is one of the most amazing teachers I have ever had the privilege to take a class from, and he and Justin are very active on the board helping people and answering tons of interesting off-topic questions too. The demonstrations he does are really cool, like making a slow-motion video of how a sound wave propagates, or another one showing how perceived gravity drops to zero in free-fall. The simulations (such as the orbit of a binary star system) are really cool too. It's a tough course even if you have more than the prereqs, but anything worth doing should be difficult. I can't recommend this course enough for anyone reading right now and considering taking the December 2013 offering. Even better, the material is spread over an extra couple of weeks this time around versus when I took it, so it should make the physics review section more manageable for people who might not have seen mechanics and E&M; in a while. But you come away from this course feeling not like you know a bunch of facts, but that you understand why so many of these facts are true. It's a lot of fun.
I said this in my review of Part II, but it applies equally well to Part I: For anyone who loves programming, this is a truly memorable course taught by one of the great heavyweights in the field. It's really something to see that such an important researcher as Sedgewick is also a phenomenal teacher. It's beautiful the way he shows the algorithms operating. Most powerpoint/pdf lectures are just frankly wastes of time to me, but Sedgewick animates his algorithms slowly and without skipping steps on his slides so that you really see how the algorithms work. It can be so hard to look at a piece of code and decipher exactly what the process is that's being executed, but the animations really help to clarify. Part I is a really awesome survey of sorting algorithms and search data structures. I loved the way Sedgewick taught how to do real, industrial-strength implementations of quicksort, for instance. I loved the discussion of the Dutch National Flag problem for partitioning. These kind of critical details are very often left out of algorithms courses where you are usually presented with a simpler quicksort that doesn't perform nearly as well. Another highlight was the bouncing balls in a container simulation (could have just as well been gas molecules in a box) using priority queues. I mean what a cool application that would be impossible to run without such an efficient data structure. The percolation project was pretty difficult, but also really interesting. Sedgewick gives such a clear discussion of union-find. My favorite project though was the one on K-d trees, which are extremely important in 3D graphics. It was by far the hardest project in either Algorithms I or II, but so satisfying once you get it up and running and meeting all performance deadlines. The only complaint I have about the course is the left leaning red-black trees. While the code is really really short for them, it's also much more clever and difficult to follow than the more straightforward CLRS implementation, and it also doesn't perform as well (which is why you see CLRS style red-black trees in the C++ STL, for instance, rather than Sedgewick's left leaning red black trees). Don't get me wrong: the left leaning red-black tree data structure is pretty cool, but I think it would be better taught as a modification of the classic RB trees. To close this review, this comment I made in the review for Part II applies equally well to Part I: Professor Kevin Wayne was EXTREMELY active on the forums. I can't count how many times I'd have a question about something and Professor Wayne would weigh in within a couple of hours to answer me. I don't know how he has the energy, but we'd be fools to not take advantage!