Back to Probabilistic Graphical Models 1: Representation

4.7

stars

1,197 ratings

•

259 reviews

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....

Jul 13, 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!!

Oct 23, 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).

Filter by:

By Jonathan H

•Jun 25, 2017

Excellent course. The video lectures are challenging (had to keep my finger on the pause key) even if you're familiar with the math, since the instructor encapsulates concepts in an amazingly concise manner. This pays off with a lot of "Aha!" moments as strong concepts are combined to create insights, especially starting around week 3. I'm already in love with this subject after 1 part

It would have been nice to have more worked homework problems, since this is a math course. But, this is not necessary to pass the class or understand the concepts. I've purchased Prof Koller's text on PGM and hope to solidify some of the intuitions I'm missing shortly.

Taking off a star because the test cases and grading software for the honors homework assessments were clearly low effort and sometimes incorrect. There were a lot of cases where functions passed all the provided and automatic test cases despite major flaws (e.g. not working for any cases besides n=1), which made it difficult to tell if things worked since the programming style is unique. The homework itself was super interesting and valuable, but I probably spent over 50% of the time fighting the grader instead of learning stuff. Given that I'm a professional programmer and completed most of the homework in 25-50% of the estimated time, I'm guessing that the average student wasted even more time with issues that are ultimately unrelated to our understanding of PGM.

By Hunter J

•Jan 12, 2017

Before I took this course I took the Stanford Machine Learning course, which I greatly enjoyed. That course allows for the learning of difficult concepts in a way that I found less painful than working through a textbook. In this course there is a lot less video content, and the coding assignments are less interesting. Expect to spend a lot of time understanding the nuances of the code that the instructional team has developed, and be prepared to really pore over the gritty aspects of Octave or MATLAB. If you're serious about this course I suggest buying the accompanying book. The slides are not easy to understand without the audio narration, which makes them difficult to review, and unlike the case in the ML course, there are not a lot of readily available open introductions written on the topics.

By Stephen A

•May 18, 2018

I really enjoyed this course. Prof Koller presents the material very well, and it's really interesting to see how probabilistic graphical model frameworks are underpinned mathematically. I thought it was a pretty tough course at points, and while the lectures are good I found having a copy of Prof Koller's textbook very useful.

I would give this course 5 stars, but I thought some of the programming assignments involved too much grappling with MATLAB rather than illuminating the principles in the lectures. Also, I think the order of the lectures may have been changed since the course was first run as there are occasional references to things that have not been covered at that point.

Overall though, very enjoyable. I'm looking forward to parts 2 and 3.

By Michael K

•Nov 14, 2016

This excellent course is exceptional in that very few MOOCs are taught at this graduate level. Others have pointed out that while this is an introductory course to Probability Graphical Models, I would say that this is still an advanced course, with lots of prerequisites. Prof. Koller is an excellent lecturer, yet moves fast, and you'll need to do reading to fill in the gaps. I haven't been able to find a good book to accompany the course, as her book is pretty dry. I strongly recommend one complete all of the Honors assignments to get a lot out of the course. The discussion boards are not so active with plenty of unanswered questions. Doing the programming assignments will greatly enhance your skills in debugging.

By george v

•Jul 07, 2017

very nice intuition from the professor Daphne Koller and "compact" in these lectures that dont exceed 15min each. really glad i did the first one, wish i did also the other two parts, certainly will when i find the time. Just as a comment, i mostly enjoyed the programming assignments. they are very well structured and in a very particular manner, which at the same time is the strong and the weak point of the assingment, since at times i undertsood something else than what the actual implementation was. anyway they were really a challenge, and whoever manages to do them should be glad with his work. Thank you prof. Koller for this course!

By Antônio H R

•Jun 20, 2018

The video lectures are really good and are useful for guiding you through Probabilistic Graphical Models book. I did not like the honor track programming exercises, however. The problems seems artificial to me and they make use of very strange data structures (probably due to the adoption of matlab as programming language). You end up wasting a lot of time with unimportant points instead of exploring ideas and getting cool results. Furthermore, I don't think the programming exercises help to familiarize the user to any of standard tools for bayesian analysis (i.e. probabilistic programming languages and so on).

By Dat V N

•Mar 28, 2018

The course helps me understand what a probabilistic graphical model is and how and why it works. One aspect I like the most about the course is the programming assignments. Those PA really make a lot of concepts clearer although sometimes you need to think carefully when the instruction is hard to follow. I think there should be more test case and expected results so that students know what is asked and to evaluate their own code. The instructor is generally clear but sometimes she goes too fast on certain concept. The course is hard but if you gives in time and effort you can complete it.

By Mehmet U

•Jul 02, 2017

Thanks for offering this course, I have learned a lot. However, the course is quite confusing. Not everything is well defined so it is hard to answer some questions. The honors programming assignments are usually confusing in this manner. If you put in the effort to understand it thou, it can be done. To be honest thou, some misunderstanding could be given to my lack of understanding the material at first. At the same time my lack of understanding is probably caused by the course material being not so well defined. Maybe it would help if one spends more time reading the text book.

By Shantanu B

•Sep 03, 2018

This course is a very essential learning step for people who want to learn and work with Baysean or Markov networks. I think that the course can be further improved by going a little slow on certain assertions or deductions which are fundamental to the subject. Those should be properly emphasized. But overall the assignments were challenging and actually made you think about the things taught in that corresponding video.

By 李俊宏

•Nov 09, 2017

This is a tough course so it was split into 3 parts. I've learned some ideas about bayesian network and markov model. The major problem about this course is the programming assignment, which is poorly maintained. Daphne Koller is very brilliant but this makes it hard for people to catch up with her, especially for people whose mother language is not English. After all, this is an interesting course!

By Zhen L

•Nov 16, 2016

The course gives an good introduction of PGM. The highlights are the well-designed quizzes and assignments. But the videos of lectures are not good enough. It's too fast and some key concepts are not clearly explained.

After looked into another course on coursera, I add a star for this....

By Vincent L

•Mar 21, 2018

Some of the examples are a bit confusing. I mostly used logic to solve these versus following a formula. Octave was fine but I didn't know how to use SAMIAM and so gave up on the coding assignments since PGMs aren't a focus area for me except for general theoretical knowledge.

By Roland R

•Dec 20, 2017

Good course. Sometimes a little bit hard to follow. For example representation of probability functions as graphs (connection between factorisation of probabilty distribution and cliques in the graph). And I'm not sure If I can apply PGMs to real world problems now.

By Hanbo L

•Apr 30, 2017

In general this is a good introductory course. You should read the book if you want more in-depth knowledge in this field. I feel that some of the concepts can be expanded a little more, like local structure in Markov model. Overall, this is a great course.

By Rick d W

•Apr 20, 2017

Everything is explained very clearly throughout the course, and the structure they use to teach the subject , from basics to advanced material, is especially helpful. Would recommend this course to anyone with an interest in probabilistic modelling.

By 邓成标

•Nov 30, 2017

The materials are very interesting, however, this professor speaks so fast that it is hard to grasp the deep theory. In overall, this course is great. And I really need to do the assignment to enhance my comprehension about the content.

By Surender K

•Nov 07, 2016

Wonderful course with great material. Wish there were more examples in the material. Nonetheless cannot complain to get this course for free with SEE material and programming assignments (need to complete yet in this session)

By Akshaya T

•Jan 16, 2018

Some tutorials need disambiguating documentation (upgrade :)) but otherwise, the course is really good. It would also help if there is a mention of what chapters to study from the book for every lesson -- in the slides.

By Rajeev B

•Dec 23, 2017

This specialization covers a lot of concepts and programming assignments which are very helpful in understanding the concepts clearly. Although, I wish there is some form of explanation for the programming assignments.

By Alain M

•Nov 03, 2018

Overall very good quality content. PAs are useful but some questions/tests leave too much to interpretation and can be frustrating for students. Audio quality for the classes could also be improved.

By Boxiao M

•Jun 28, 2017

The lecture was a bit too compact and unsystematic. However, if you also do a lot of reading of the textbook, you can learn a lot. Besides, the Quiz and Programming task are of high qualities.

By Shawn C

•Nov 05, 2016

The course is great with plenty of knowledge. A little defect is about description about assignment. As the forum discussed, several quizzes may confusing.

By Hilmi E

•Feb 16, 2020

I really enjoyed attending this course. It is foundational material for anyone who wants to use graphical models for inference and decision making..

By Roman S

•Mar 20, 2018

A good introduction to PGM, from very basic concepts to some move in-depth features. A big disadvantage is Matlab/Octave programming assignments.

By serge s

•Oct 18, 2016

Thanks to this course, Probabilistic Graphical Models are not anymore an esoteric subject! I am really looking for the second part of the course.

- AI for Everyone
- Introduction to TensorFlow
- Neural Networks and Deep Learning
- Algorithms, Part 1
- Algorithms, Part 2
- Machine Learning
- Machine Learning with Python
- Machine Learning Using Sas Viya
- R Programming
- Intro to Programming with Matlab
- Data Analysis with Python
- AWS Fundamentals: Going Cloud Native
- Google Cloud Platform Fundamentals
- Site Reliability Engineering
- Speak English Professionally
- The Science of Well Being
- Learning How to Learn
- Financial Markets
- Hypothesis Testing in Public Health
- Foundations of Everyday Leadership

- Deep Learning
- Python for Everybody
- Data Science
- Applied Data Science with Python
- Business Foundations
- Architecting with Google Cloud Platform
- Data Engineering on Google Cloud Platform
- Excel to MySQL
- Advanced Machine Learning
- Mathematics for Machine Learning
- Self-Driving Cars
- Blockchain Revolution for the Enterprise
- Business Analytics
- Excel Skills for Business
- Digital Marketing
- Statistical Analysis with R for Public Health
- Fundamentals of Immunology
- Anatomy
- Managing Innovation and Design Thinking
- Foundations of Positive Psychology