Ok

ok

Ex

ok

.

.

K

I

E

G

I

g

The instructors a fantastic comedic duo! I had numerous laugh out loud moments while getting a much better handle on machine learning. The instruction is clear and the homeworks are always interesting and well thought out.

I really only have one complaint: Graphlab. Graphlab is a fantastic piece of software that makes data science easy. I was amazed at how much harder it was for me to complete the home works with scikit and pandas. Unfortunately, it's very unlikely that any future employer will be using Graphlab. It's a great intro for noobs (like me), but it's very expensive and can be replicated with open source packages. As such, I had to spend a lot of time trying to figure out on my own how to perform graph lab operations in the open source world. Avoiding this kind of slow, frustrating, stack exchange binging, 50 tab opening, nonsense is one of reasons I take MOOCs in the first place. I hope a future version sets up the bones of what open source packages can be used to solve each problem set.

In summary, I really enjoyed this class. I was excited to work on it. I learned things that made me more productive at work. I felt the joy of learning. I was disappointed by the insistence on Graphlab but if making the sales case for it is what motivated Carlos and Emily to put the class together, it's worth the frustration. I'm extremely grateful to both of them for putting this together.

This is the introductory course to a series of courses designed to teach machine learning through case studies. The introductory course provided a nice overview of key machine learning techniques in the context of common applications of machine learning (case studies). The concepts are taught in Python using the high-level DATO GraphLab libraries, so the course provides a good introduction to using the machine learning techniques that are provided in DATA GraphLab rather than on learning the mathematics behind the techniques.

I am investigating the capabilities and advantages and disadvantages of existing frameworks such as Pandas and DATO GraphLab, so this course met my needs, but it would probably not be as useful to data scientists who are interested in learning the underlying principles, mathematics and numerical methods used in machine learning.

I would have given the course a rating of five IF it had done a better job of describing the fundamental principles of machine learning and if it had made it easier to use a non-commercial framework such as Pandas.

I'm planning to take the next course in the series in the hope that it will provide more in-depth coverage of the fundamental principles of machine learning than the introductory course.

First parts of the course was great, week 6 was not the same quality of as the week 1-3, week 4-5 was ok.

Some quiz questions was verry hard, both good and bad, I had trouble understanding the questions and also why an answers was correct. I´m working as a QA manager/test lead and in my role I also have access to data from our production system, so I was able to use the predicting house prices model as a template and made a test on my own data, turned out to be wokring quite well. I also made a program to fetch incoming production cases and converted these to text blob, then did the TF-IDF classification nearest neighbor using Graphlab, that turned out great. So I can actually show of a POC on a classification module forour incoming Cases (in this case it is PDFs with import export data such as invoices, CMR´s etc.), the incoming data would then be able to find similar documents and clusters/distances. I was able to do all that stuff just from a couple of hours of learning, so great tool for a beginner. I haven´t seen any prices on licence costs for graphlab, but I guess that if I was going to use this at work in real, I would look into both open source and licenced software to do the classification and linear regressions.

The course approach is excellent and eye-opening for someone who studied computer science 20+ years ago. The only real reason to reduce by one star is that the setup of the software didn't work as advertised. In a way, this is to be expected: it is like giving turn by turn directions without having a map, and running into a road block (in this case: the instructions didn't work exactly as advertised for one step, probably due to the software versions moving on). The discussion boards didn't really help, as they just had other students with the same issues and a mentor saying he had created a virtual machine somewhere which I didn't want to use. As a result, I had to quickly learn a lot more than I thought about pip and python and conda - discovering the roadmap, as it were. It was worth doing, though, because once I worked through all that, the learning material and exercises all worked perfectly and as advertised - and quite doable in the advertised time.

The course material is sound and well presented. The speakers are articulate and convey the material well. The lectures cite graphplotlib, but the programming assignments utilize Turicreate, which was last released (6.4.1) in September 2020 and is not supported on python releases after 3.8. Programming examples in the videos vary slightly, but not in a blocking way. Turicreate might have been a reasonable choice when the course was written but is a bit outdated and introduces additional complexity in setting up a local environment relative to current versions of Ubuntu/Python. One additional challenge is that there is no course specific Discussion in the community and it seemed like I was on my own in figuring out syntactical/semantical errors when trying to use some features of the toolkit. I started the specialization, and will likely complete it but had I known this I would have likely chosen another offering on this topic.

I really enjoyed this course as one that's new to Python and Machine Learning. The intuitive approach of teaching the overview was very helpful to me. It is exactly what I expected from the course.

The reason I do not give this course five stars is the use of GraphLab. I lost the first four days of my 1-week free trial week trying to install in on my Mac. After failing several times using the Mac installer, I finally noticed that it has not been updated since the course was taught in 2015. I was ready to quit Coursera before my 1-week free trial ended. Thankfully someone quickly answered my question in the forum and provided the steps she used to successfully install GraphLab. I'm also concerned that I will have to pay for GraphLab after a year if I'm not taking another class. Outside of maybe getting a machine learning job at Apple, I wonder what's the use of learning GraphLab instead of other tools.

Overall, this is an interesting course that covers a lot of different topics in a compact manner. I also appreciated the combination of broad "pictures" about the different ML algorithms and ML in general.

I only gave four stars and not five because of the lab parts. Even if the topics were clearly of interest, they are really on the easy side. On top of that, using graphlab sometimes hiders the comprehension. It would have been nice to go more into the details of (at least some simple) algorithms. There is a balance between user experience (how easy it is to use powerful algorithms such as deep learning and see in practice their results) and learning experience (how much I do really myself). I feel that the balance is too much geared towards user experience. I would have been rewarding to implement some simple algorithms ourselves (such as basic regression, popularity classifier, ...).

A phenomenal, deliberately shallow introduction to Machine Learning. This course definitely cannot standalone but serves as a great start to the specialization.

There was a great deal of sensitivity in the course due to the use of Dato Graphlab, particularly since one of the instructors is the CEO of Dato and the software is not free for commercial use. This was exacerbated because the color scheme of the presentation slides matches the color scheme of the Dato website.

However putting this aside the material is very high quality. The use of case studies is inspired and helps students focus on when to use particular ML techniques. Although the material is a little shallow (deliberately) you get a good high-level feel for what techniques are applicable in specific situations. However the course can't really standalone, and it's intended as part of the specialization.

A superficial but good intro to machine learning which only requires a bare minimum of mathematical background, making it accessible to almost everyone! professors are funny and super clear, I recommend it to everyone who wants to have a broad understanding of what machine learning is about, being able to understand most of the news and discussions about the subject (except those about neural networks and reinforcement learning), with a small weekly effort. It is also a good way of starting before going into more serious stuff and coding (as I personally did). In my case, this served to understand the principles and the possible applications of machine learning, as well as getting a first touch with the coding side. Afterwards, I decided to deepen some of the specific algorithms and techniques with scientific articles and coding projects.

I think the course is very good. I do not give the fifth start because there are unnecessary elements that the trainers do, maybe because it is fun for them, that do not add to improve the learning process. Personal opinion like if a house has many bathrooms and make laugh of that or trying to be fun with their personal life. All of that is unnecessary because knowing personal details of their life, did not help me to understand and learn better. It was more like a distraction.

The other thing that I still do not understand is, why to use Graphlab, if it is a tool only for academic purposes. If I cannot use it for commercial purposes, why to present the course with that, even if it is very powerful. It will have no use for the people that are in the commercial world and we will have to learn again how to do with other tools.

This course was simple. Since it is intended as an overview of few machine learning techniques used to build computer based rational agents it has completely fulfilled my curiosity about what intelligent applications can be built using machine learning algorithms. Professors Emily Fox and Carlos Guestrin did an excellent job of presenting the course in an engaging way and with enough hands-on exercises to better understand the lessons.

I am excited to take the next specialization courses to better understand the inner workings of each algorithm, and perhaps manage to understand why these algorithms work the way they do, specially the Deep Learning algorithms.

Thank you Coursera for the opportunity and thank the Professors Emily and Carlos for such a valuable gift of sharing knowledge for free.

This is a very good introduction to the more popular types of machine learning methods implemented along with business use-cases to help give context to what you are applying. The assignments were not always clear, but thankfully the discussion boards had a lot of helpful information. I would have given it a 5, but this course is heavily crafted around the use of graphlab, which is definitely an easier way to quickly access pre-built models and libraries, the teachers in this course also have a stake in the software as they helped develop it and distribute it. It seems like a bit of a conflict of interest, and at times the lessons seem to "sell" the product, but hopefully the additional courses will provide a more raw look into the nuts and bolts of building these algorithms.

Very well structured, I'm really looking forwards to the next courses.

I like how they always start from a real-life, down-to-earth example and work their way up to the technical bits. Everything is clear and explained in detail, and often multiple solutions are proposed: e.g. 'we do A, but this has the problem blah, so we do B..' super good, comparison is the key for understanding. Python notebook is an amazing pedagogical tool, and I really appreciate how the exercises are designed: first go through pre-made code, then tinker to get a few variations.

I would like to see a bit more of the 'algorithms' (e.g. could outliers make centroids drift away in kNN? what is the computational complexity of the various functions we're using?) but it's just a minor remark.

Way to go guys!