No python..

THANK YOU.

Too easy

good

Good

cool

Good

good

V

The content of this course is interesting, I liked the examples, and the material gave an interesting overview of different aspects of machine learning. From that perspective, the course is as advertised. But, where this course goes wrong is value for money - it is very superficial and not worth what is charged.

As noted by others, this is not a course for learning so much as an advertisement for the instructor's own pay software and their other Coursera courses. I'm not against that per say if it was entirely free, but charging for an advertisement is ridiculous. In my case I thankfully started with the free model so I didn't lose out, but I could see others being dissapointed. I strongly recommend starting the material with a free signup and only pay if you really want the extra grading.

My other main problem was with the pace and detail in the course. I would have liked more detail, but I recognize this was intended to be a high level view so I'll live with that level of detail. The material covered, however, does not need 6 weeks worth of lectures. This course could be ~1/2 as long, cover the same material, and be a MUCH better course.

Other small problems include some poorly edit videos (there are a lot of examples of simple stumbling in the videos that should have meant they do another take), very short videos (maybe a person preference, but the number of <2 minute videos here is annoying, especially when there's a 5-second standard video at the start and end of all videos). All in all, there's just a lot of wasted time.

When signing up for this course I was really excited for the entire specialization - now, not so much. I'll probably try the second course in the series (for free to start) to see if things improve, but ironically this advertisement video has if anything turned me off their other products.

Con:

(1) I feel I spent most of the time learning graphlab. Suggest replace it with standard Python as the standard tool for this class. Provide any needed additional code in standard Python.

(2) Course is better in the front end than in the back end.

(3) Week #6 is significantly more involved than previous weeks. Suggest divide Week 6 into two sessions: Neural Network and Nearest Neighbor applying neural network results (ImageNet 2012 was mentioned and not explained. Therefore the Nearest Neighbor homework assignment from the student's perspective does not have much to do with neural network other than using the results from ImageNet 2012, which was not explained in any detail anyway). This will allow more time to delve into the forward and backward propagation which should have been explained in more details.

(4) Home assignments are not best worded, especially homework assignment for Week 6. Suggest reword in shorter statements that are more to the point.

(5) Programming presentation and assignments can seem like exercise in graphlab and SFrame functions rather than machine learning.

PRO:

(1) Class presentation by Professor Fox on recommender system is detailed and clear.

(2) Classifier block diagram shown by Professor Guestrin is good, clearly distinguishing training the classifier and the subsequent use of the classification (prediction).

(3) Neural network quiz in Week 6 is excellent. It drills down on the multi-dimensional space that neural network is particularly good for.

Even though the course discusses relevant topics, the level is extremely low: The lab sessions were easily solved applying copy-paste code from the provided notebooks, with minor adaptions. Moreover, 8/10 questions in the lab sessions were not related to machine learning at all, but simply looping over data and counting or similar. The intro video and course introduction strongly suggested using deep learning in the course: did not happen. We trained k-means on pre-computed features which happened to come from from a deep learning network (not sure which one, inception? I didn't even watch the lectures here from disappointment). That is not deep learning, it just shows you how well deep learning can work.

Graphlab is a mature framework, I guess, but it's commercial and scikit-learn is better imho (and free!).

If you wish to learn machine learning, take the Stanford course on Machine Learning for Andrew Ng. This course is in MATLAB, not ideal for machine learning, but adequate for a better understanding of intelligent system implementations.

Maybe the course is OK if you're a beginner in machine learning, but not good.

I leave 2 stars as I learned a lot of new information and methods, and the theory and math behind them.

You will learn about Data Science and Machine Learning, but not much about Python.

The course is pretty much abandoned and outdated. Sframes and Turicreate packages (instructor's creations) are used instead of more universal packages. Installation in the beginning took some time and research. Many of the assignments have errors and bugs in the code that have not been updated. Forum assistance is abysmal for clarification or deeper questions. Many links are dead.

There are many times in the lectures where the instructors are writing several sentences in their handwriting on their notes instead of having the text ready to appear.

I would suggest using this course and series as a supplement to other information one as learned, not as an introduction for initial understanding. I found myself frustrated too many times.

Course doesn't do nearly enough to bring you up to speed on using mathlab or iPython notebook. I am currently learning to program python and a lot of this stuff was well above my head.

The quizzes and assignments do very little to reinforce the work, and often come down to trial and error. I wanted to learn the mechanics of machine learning from this course, but it is too complex, and presented in such an arcane manner to serve as an introduction, but doesn't go deep enough to really teach anything useful. I'd suggest you look at Wikipedia or YouTube for better classes.

I'd like to draw special attention to the quizzes, as often they're on trivia from the lectures and not reflective of the actual nuts and bolts of working with machine learning. They, as with the projects, I found to be a massive waste of my time.

The teachers are easy to like, but the course content is very lightweight and will mostly teach you terminology with no real understanding.

The worst part was the assignments, which could all be solved by a little copy/paste: I didn't learn anything useful by doing them. All the actual algorithms were supplied in a separate module. More than that, many of the suggested solutions were bad coding (like collapsing 50% of the data before training, or writing sixteen special cases rather than a general function) or pointless (like training a linear classifier on pixel data).

There are better courses out there.

Good course! The only major drawback is the requirement of Graphlab, which doesnt allow the student to fully understand the applications using real world software. Just recently, Dato (the company that owns graphlab) was purchased by Apple, and you can no longer buy a commercial licence to the software. Despite this, users cannot use Graphlab for commercial purposes, therefore rendering the software completely impractical for professionals. The specialization is designed to help you get a job (see capstone) yet the software currently in place is limiting.

I have no doubt teachers are excelent professionals in the area, as well as great machine learning enthusiasts. However, I did not like the fact that you get limited to learn how to use a paid and (very!) expensive platform, mostly because there are many other free packages available for machine learning. Ok, the platform offered makes things easier, but if you really want to learn machine learning, you can not be limited to a platform, acting as a robot just using pre-written functions in a black box.

2 stars because the theoretical part is ok but programming assignments are waste of time. I'm not here (and paid) to be trained to use something the instructor is trying to SELL, nor will I ever recommend this product for commercial use. I will switch to other "not recommended" packages in the later parts of this specialization.

They should put the disclaimer for Graphlab Create in the specialization page so people can be aware of this.

Besides, the sound of that Giraffe toy is really, really annoying.

It really just touches a lot on different machine learning techniques and really just sets the stage for the higher courses. Unfortunately some of the chapters (especially deep learning) are so brief that it is really frustrating trying to complete the quiz and assignment. Also the course doesn't use open source tools but a trial version of a pretty expensive library.

Poorly executed. Constant differences in data. tiresome to watch two supposedly very intelligent instructors amuse themselves by saying Bro and Dude. The use og graphlab is unnecessary and adds a layer of complication which adds no future value to your toolkit. Probably a lot of better executed Machine Learning courses out there

I like the case-based approach--this course gives a nice albeit shallow overview. I don't like that one professor uses this course to push his startup by asking students to use graphlab. A more commonly used library would have been a much better choice. Parts of the course feel like a "Getting started with Graphlab" tutorial.

Having done some other machine learning MOOCS , this course seemed rather basic to me and did not enjoy too much using non open-source software for the programming assignments. The material is nice, In this sense, I would have expected to 'default' to sci-kit learn and offer using graphlab create as optional.

While it was good for learning concepts I had real trouble with graphlab. Installation of graphlab never worked on my machine. I had to install VM just for being able to use graphlab. I really wish they had opted for more open source, free options or at least used ince such library along with graphlab.

There are big problems in this course, like the installation process should be given in a more specific and vivid way so that I would not have spent three days on it being a windows user...(update: still can't access jupyter notebook after trying installing ubuntu, vmware workstation, filezilla).

Too much dependency on Graphlab package is bad. If we are learning the concept, we should reduce the size of the sample data. We should be using generic open packages, so that our learning can be easily demonstrated anywhere (especially interviews), and not dependent on graphlab.