Chevron Left
Back to Probabilistic Graphical Models 3: Learning

Learner Reviews & Feedback for Probabilistic Graphical Models 3: Learning by Stanford University

4.6
stars
298 ratings

About the Course

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 third in a sequence of three. Following the first course, which focused on representation, and the second, which focused on inference, this course addresses the question of learning: how a PGM can be learned from a data set of examples. The course discusses the key problems of parameter estimation in both directed and undirected models, as well as the structure learning task for directed models. The (highly recommended) honors track contains two hands-on programming assignments, in which key routines of two commonly used learning algorithms are implemented and applied to a real-world problem....

Top reviews

SP

Oct 11, 2020

An amazing course! The assignments and quizzes can be insanely difficult espceially towards the conclusion.. Requires textbook reading and relistening to lectures to gather the nuances.

OD

Jan 29, 2018

very good course for PGM learning and concept for machine learning programming. Just some description for quiz of final exam is somehow unclear, which lead to a little bit confusing.

Filter by:

1 - 25 of 53 Reviews for Probabilistic Graphical Models 3: Learning

By Akshaya T

•

Mar 14, 2019

I was very lost with the different depths of lectures and assignments in this part of the course. I felt that some places were super involved mathematically and was trying to understand its implication. In other places it felt like a lot of fluff. I would recommend this only if you have taken the other 2 parts. Also Prof. Koller's lectures are quite confounding and monotonous in these more than the other lectures.

By A M

•

Jun 17, 2019

Great lectures and terrible assignments. Forum is not helpful at all. In fact, the forum is dead and tutors do not exist. Programming assignments have too many errors which are known within the forum for 4 years but no one is fixing these mistakes. All in all, the topic is highly interesting but the implementation is deficient

By Ahmed S

•

Sep 22, 2017

Pros:

The course covers a highly important relatively large set of topics. If you get the content and managed to pass the quizzes and assignments, you're good to go with PGMs.

Cons:

The course is quite old, with no support from neither TAs nor instructors. The material isn't updated to match a specialization (even the assignment numbers are old, some test cases aren't updated and the course content and assignments are quite dependent).

By Maxim V

•

Apr 30, 2020

Great course, especially the programming assignments. Textbook is pretty much necessary for some quizzes, definitely for the final one.

By Jesus I G R

•

May 30, 2020

1) The fórums need better assistance.

2) If we could submit Python code por the homework assignments, that would be much better for me.

By Michel S

•

Jul 14, 2018

Good course, but the material really needs a refresh!

By Rohan M

•

Dec 5, 2019

Some excellent materials and homeworks, but poor teaching support and poor value for money.

By Dat N

•

Nov 14, 2019

The course really helps me understand a lot of things about learning graphical model, from estimating parameters for Bayesian Network, Markov Random Field, CRF, to learning graph structure from data and using EM algorithms to learn when there is missing data. It also gives many guidelines about the process of machine learning in general. I found the programming assignment more challenging than the first 2 parts but at the same time they are very enlightening when all the pieces beautifully fit together. In general, it was a fun, challenging and enlightening learning experience. I want to thank the course instructor and staffs who made this great course possible.

By Lik M C

•

Feb 23, 2019

A great course! Learned a lot. Especially the assignments are excellent! Thanks a lot.

By Shawn

•

Aug 20, 2020

The course content is great, prof Koller will introduce to you the most useful techniques in PGM and demonstrate algorithms with good examples. This is far more efficient compared to reading a book filled with mathematical expressions, especially if you are new to this area like me

However, I found the ppt could surely need some polishing. During the course 1/2/3, a lot of times you will find that prof Koller speaks about many details and crucial facts that are not even shown in the ppt! Although she wrote down notes sometimes but the notes were hard to recognize (might need a OCR lol), so this is not good if you take screenshot and want to review later.

The 24h cooldown for the final course exam should be reduced at least, if not removed. 4 to 6h is sufficient for a good student to review problem and make correction. And also keep in mind that for the final exams, you don't get to see your answer or any hint about your error once you submit, thus quite challenging and sometimes frustrating, but could be rewarding once you pass.

Among other things, some lectures have bad audio quality, CC are incomplete, and the discussion forums are not very active.

Anyway, if you're about to take the final part of this specialization, good luck!

By Diogo P

•

Nov 15, 2017

Just completed the 3 course specialization. If you're interested (and already have some background) in Machine Learning, this specialization is totally worth it. However, if you have trouble solving any of the quizzes or assignments, do not expect to have any kind of support from the TAs. They simply do not respond to any post in the forum, even if it is related with any bug in the programming assignments source code.

By Rishabh G

•

Jun 3, 2020

Old but worth it. Professor gives a very clear and intuitive explanation of things. Although the course is not as tough as would have been in a real Stanford classroom. But good for beginners for getting started in PGMs. And to see how cool they are. It intrigued my interest more in this field.

By Marcelo B

•

Mar 6, 2021

Although the contents and the way Daphne explains the subject are of top quality, the rest of the specialization leaves a deep frustrating feeling. There is no TA present at all during the courses. Since the quizzes and final exams are dubious, sometimes pedants, and written in an extremely confusing fashion, you end up guessing instead of applying what you have learned. The book is essential, not a recommendation, and needs to be carefully read. Moreover, PAs are not clear, sometimes as a chunk of some other larger PA, the code is full of bugs so that sometimes only Matlab (nop, no Python, sorry) works, and sometimes you have to look into the net to see how to correct bugs to be able to submit the code. Again, no TAs, so all fall into the blogs, and forums. This is very, very frustrating. The final exam can be re-taken in intervals of 24 Hs. So if you happen to start three days before the end of the course you may find yourself paying again to just take the final exam. An interval of 4-8 Hs may be sufficient since the lectures are not so long. Since the confusion in the questions of the final exam is huge, almost sure you will have to re-try the exams at least twice. I think the intention was to make the course intensive and difficult, but the way it was chosen, transforms the course into an epic failure. Summary: if you are looking to gather knowledge, stay away from this specialization. There are plenty of free courses with deep explanations, addressing modern techniques, and correct code PAs, under the same limitations: you are alone, no TAs or colleagues to be asked. (e.g., https://www.cs.cmu.edu/~epxing/Class/10708-20/lectures.html, by March 2021). If it happens that the specialization is paid by the company you are working for, then go ahead, but keep in mind that you are alone. If it is no, do not waste time and money. I have finished the specialization with honors (almost 100%). Still, I am deeply disappointed. The two stars are given due to Daphne.

By Antônio H R

•

Nov 6, 2018

Bad choice of content. Focus too much on the specific case of table CPDs, missing the big picture.

By Sergey S

•

Sep 24, 2020

Always 5 stars!!! This module is dedicated mostly to HMM and learning latent variables. I could never imagine, that explanation of a such hard topics can be so gentle and simple! The home assignments here are very... extremely juicy: every assignment gave me 4-5 punches! I really didn't know how to complete them. It forced me to review lectures 3-10 times, dive deep into theory details and to reassemble all previous material together again. And it was just WOW! Man starts to see much more after home assignment completion. And it is what makes the whole course extremely well done! (The all 3 modules took me around 8 month to finish). Many thanks!

By Alfred D

•

Aug 13, 2020

Very good course , but also very difficult subject ; There was a lot of effort to sit through some of the lecture videos averaging >23 mins and also maintain comprehension and relativity. The notes could have been more better devised especially for the exams ; I had to do a lot of external reading to get through the exams and some questions are still ambiguous to the possible solutions. A big thank you to Prof Daphe Koller and I saw this as an opportunity to have class time with a Prof of her calibre, which I may never get . All in all it was hard but in the end the revelation of learning something so complex was well worth the blood and sweat

By Shi Y

•

Jan 19, 2019

I love this course! It's very difficult but worthy. If you are looking for the state-of-the-art AI techniques, PGM doesn't seem to be your best choice. It's some kind of old fashion compared to DL. I learned a lot about the probability theory through all three courses, and I get better understanding with CRF and HMM. Seriously, it's not a course that will improve your skills or guarantee your successful immediately in ML fields, but a course that can shape your thoughts, help you think out of box. So if you don't like the black-box in DL, PGM will offer you another brand new perspective to understand this uncertain world.

By Chan-Se-Yeun

•

Feb 21, 2018

Yeah! I managed to finish PGM. I feel ready to explore further. PGM 3 is really helpful. Although many details are not fully discussed, some important intuitions are well illustrated, like EM algorithm and its modification in case of incomplete data. Also, the way the teacher teach set an good example for me to learn to demonstrate complicated things in an easy and vivid way. Thank you so much!

By Rishi C

•

Jun 4, 2018

The course facilitates learning - and reinforces acquired knowledge through the simple principle of honest effort: students are not given all the answers... but they are 'nudged' in the right direction & guided towards fruitful questions; in a way, it's the perfect course!

By Musalula S

•

Aug 25, 2018

The course is very involved but Daphne makes its palatable. The course open a new world of new possibilities where one can apply PGMs to get concrete understanding of relationships between events and phenomena in any discipline; from social sciences to natural sciences.

By Joseph W

•

Jan 9, 2020

Great course, very dense but informational. Took a lot of time for content to sink in and I had to review it several times, but now I feel confident in my ability to learn structure/parameters in graphical models.

By Jaime A C

•

Nov 15, 2022

It's an amazing course that has allowed me to connect a lot of different areas of bayesian statistics, machine learning and probability theory. The material is top notch and challenging, but very rewarding.

By satish p

•

Oct 12, 2020

An amazing course! The assignments and quizzes can be insanely difficult espceially towards the conclusion.. Requires textbook reading and relistening to lectures to gather the nuances.

By Orlando D

•

Jan 30, 2018

very good course for PGM learning and concept for machine learning programming. Just some description for quiz of final exam is somehow unclear, which lead to a little bit confusing.

By Henry H

•

Feb 13, 2017

Great course! Very informative course videos and challenging yet rewarding programming assignments. Hope that the mentors can be more helpful in timely responding for questions.