返回到 Probabilistic Graphical Models 1: Representation

星

1,358 个评分

•

302 条评论

Probabilistic graphical models (PGMs) are a rich framework for encoding probability distributions over complex domains: joint (multivariate) distributions over large numbers of random variables that interact with each other. These representations sit at the intersection of statistics and computer science, relying on concepts from probability theory, graph algorithms, machine learning, and more. They are the basis for the state-of-the-art methods in a wide variety of applications, such as medical diagnosis, image understanding, speech recognition, natural language processing, and many, many more. They are also a foundational tool in formulating many machine learning problems.
This course is the first in a sequence of three. It describes the two basic PGM representations: Bayesian Networks, which rely on a directed graph; and Markov networks, which use an undirected graph. The course discusses both the theoretical properties of these representations as well as their use in practice. The (highly recommended) honors track contains several hands-on assignments on how to represent some real-world problems. The course also presents some important extensions beyond the basic PGM representation, which allow more complex models to be encoded compactly....

ST

Jul 12, 2017

Prof. Koller did a great job communicating difficult material in an accessible manner. Thanks to her for starting Coursera and offering this advanced course so that we can all learn...Kudos!!

CM

Oct 22, 2017

The course was deep, and well-taught. This is not a spoon-feeding course like some others. The only downside were some "mechanical" problems (e.g. code submission didn't work for me).

筛选依据：

创建者 Rishabh G

•May 11, 2020

Great course. Explained in a straightforward manner.

创建者 Lorenzo B

•Jan 19, 2019

The course contents are presented very clearly. Difficult ideas are conveyed in a precise and convincing way. Despite this, the global structure is not presented very clearly, and the quality of some course material is not excellent. In particular, I didn't find the optional programming assignments particularly interesting, and the code/questions contained more than one bug. Also, the quality of video/sound is quite poor, and varies a lot from course to course.

创建者 Michael G

•Feb 5, 2017

The support by the mentors could be much better. Because of the missing support I was not able to solve the assignments under Windows with Octave. I had to buy Matlab. (-2)

It seems to me that the course is very difficult to complete without additional sources. (-1)

创建者 Benjamin B

•Apr 12, 2018

Did not like how the concepts were introduced, it felt like learning theory for the sake of theory.

创建者 Sumod K M

•May 6, 2019

The course contents and presentation is of very high quality. The assignments and quizzes are both challenging and very rewarding. The only minor qualm is that the programming assignment grader seems to have few issues. For one, MATLAB indexing is really hard to work with. Secondly, it doesn't test the answers fully in some cases. Like the case of OptimizeWithJointUtility, OptimizeLinearExpectations. My codes passed the grader but I was splitting to hair to figure out why my answers to quiz questions corresponding to programming assignment were wrong. Turned out that my code was incorrect for the two programming assignments and that was causing issues. Otherwise, really nice course. Thank you :).

创建者 Ka L K

•Mar 27, 2017

A five stars course. Prof. Koller is an outstanding scientists in this field. The first part just introduce you two basic frames of graphical models. So go further into second part is necessary if you want to have a bigger picture. The whole course is an introduction to the book - Probabilistic Graphical Models of Prof. Koller, so buying her book is also highly recommended. This course is supposed to be hard, so you should expect a steep learning curve. But all the efforts you made are worthy. I suggest coursera will consider put more challenging exercises in order to extent the concentration. Finally, a highly respect to Prof. Koller who provide the course in such a theoretical depth.

创建者 Marcelo B

•Nov 25, 2020

The course is very well organized and good leveled. The contents you get from the videos need to be completed/understood with the book. This makes this course a hard one, but very enjoyable. Having said that, I would suggest some improvements, if I am allowed. The first one is to update the course material to reflect the current scope of machine learning (e.g., Deep Learning). The second one is to include the option to code in Python. The last one is related to the final grade. I believe that giving the 24hs submission option is exaggerated. I really enjoyed the course and got a vision on PGM that will allow me to apply them in my work.

创建者 Sha L

•Apr 19, 2017

it's really hard course for me but after completing and see the certificate I feel so good about it. Yesterday someone asked a question regarding conditional independence. I remember before I took the course I've spent quite some time understanding it, just like him. But yesterday I didn't event think about it and gave him the right answer using "active trail" and "D-separation" concept. That's how powerful this course can be.

I didn't work on the honor track though because I'm currently short of time. But I think I will come back and taking the other 2 courses in this series.

创建者 Blake B

•May 21, 2017

Awesome intro to graphical models, and the exercises really emphasize understanding and proceed at what seems like the appropriate pace. Challenging for sure, you need to want to learn this stuff. Only downside is I'm not a fan of using octave/matlab--really wish this could be rebuilt using python for all the exercises. I've probably spent 60% of my time devoted to this course on getting that setup working and wrestling with telling the computer to do what I want in an unpopular language--at least, unpopular out in the world outside of academia.

创建者 Chan-Se-Yeun

•Jan 7, 2018

This course is quite interesting not that easy. It helps me understand Markov network. The questions within the video are very helpful. It helps me check out some essential concepts and details. What's more, I'm fascinated by the teacher's voice and her teaching style, though detailed reading is required off class to gain comprehensive understanding. This is the first time I take online course in courser, and it's fun. I think I'll keep on learning the rest 2 courses of this series.

创建者 Haowen C

•Sep 1, 2017

Excellent course for picking out just the critical portions of the Koller & Friedman book (which is over 1000 pages long, forget about reading it cover to cover for self study). Don't skip the programming assignments, they're very important for solidifying your understanding. You'll spend at least 75% of the time fussing over the somewhat arbitrary and baroque data structures used to represent factors and CPDs in this course, but at the end it's worth the frustration.

创建者 Dawood A C

•Oct 25, 2016

The course was very fruitful. It is was not that easy of course, I think it is one of the most difficult courses on Coursera but it deserves to try it once, twice and as many as you can until you understand the idea behind the course. The exams and the honor assignments were so tricky and not that easy to solve. If you don't have a probabilistic background, I think first better for you to take a course in data analysis and probability.

创建者 Wenjun W

•May 21, 2017

Awesome class, the content is not too easy as most online courses. Still the instructor states the concepts clearly and the assignments aligns very well with the content to help me deepen my understanding of the concepts. The assignments are meaningful and challenging, finishing them gave me a great sense of achievement!!

It would be better if the examples in the classes could incorporate some industry applications.

创建者 Rishi C

•Jan 29, 2018

Perhaps the best introduction to AI/ML - especially for those who think "the future ain't what it used to be"; the mathematical techniques covered by the course form a toolkit which can be easily thought of as "core", i.e. a locus of strength which enables a wide universe of thinking about complex problems (many of which were correctly not thought to be tractable in practice until very recently!)...

创建者 Dimitrios K

•Oct 31, 2016

So happy to complete this one. It was tough - especially the programming exercises and mainly due to high degree of vague-ness and un-expressiveness of matlab/octave in contrast to e.g. Python or Scala. samiam was unexpectedly handy and usable. Very nice and educational piece of software. Excellent course - it's incredible how many Machine Learning models are expressed under the umbrella of PGMs.

创建者 ivan v

•Jul 31, 2017

Excellent introduction which covers a wide range of PGM related topics. I really liked programming assignments. They are not too difficult but extremely instructive.

Word of advice: although programming assignments are not mandatory, dare not to skip them. You will be missing an excellent learning experience.

Another useful advice: lectures are self-contained but reading the book helps a lot.

创建者 Meysam G

•Sep 12, 2019

I had actually read the David Barber book before I took this course. The course provides a deep insight to the PGMs which is necessary if one wants to utilize it in real applications or as in my case in research works. Moreover, the language of the instructor is comfortably plain, especially when it comes to explaining somewhat complicated concepts. In general, it is highly recommended.

创建者 Gautam K

•Oct 17, 2016

This course probably the only best of class course available online. Prof Daphne Koller is one of the very few authority on this subject. I am glad to sign up this course and after completing gave me a great satisfaction learning Graphical Model. I also purchased the book written by Prof. Koller and Prof Friedman and I am going to continue my study on this subject.

创建者 Diogo P

•Oct 11, 2017

Great course. The lectures are rather clear and the assignments are very insightful. It takes some time to complete, mostly if you are interested in doing the Honor programming assignments (and you really should be, because these are demanding but also very useful). Previous knowledge on basic probability theory and machine learning is highly recommended.

创建者 SIYI Y

•Nov 3, 2016

This is definitely a good course. The honors assignments are interesting, which instruct you to implement graphical models from scratch to solve problems in real world using Matlab or Octave. This helps me understand the theory part better and allows me to have better sense how they can work practically applications.

创建者 Siwei G

•Jun 7, 2017

It's a great class. A lot people may complain that there should be more details. Well, this course may not hold your hands all the way to the end, but it covers enough to get you started to learn independently. It is a graduate level class, and it should be designed in this way. 5 star for the wonderful content.

创建者 Jaewoo S

•Apr 26, 2020

Wow, it was a hard course. And it is usually true for hard courses, I really learned a lot. I truly recommend to solve all honors contents to get thorough understanding. Meanwhile, some programming assignment contents need to be either fixed or improved. There have been many discussions in the discussion forum.

创建者 Eric S

•Feb 1, 2018

A very in depth course on PGNs. You definitely need some background in math and a willingness to invest a lot of time into the course. Of most value to me were the programming exercises. They are in Octave as this is one of the earliest Coursera courses, but it is worth exploring the provided implementations.

创建者 Douglas G

•Oct 24, 2016

This course is very help for who have to study anything the respect of machine learning example, which is a thing much used in every day and in the new context of new industries 4.0, and the studies of probabilistcs graphical can help who need to develop new programs each times more efectiviness and best.

创建者 Venkateshwaralu S

•Oct 25, 2016

I loved every minute of this course. I believe I can now understand those gory details of representing an algorithm and comfortably take on challenges that require construction and representation of a functional domain. On a different note, nurtured a new found respect for the graph data structure!