In this post, I would briefly discuss the references that helped me understand important concepts in the broad field of Probabilistic Graphical Models. Given the sheer amount of topics that come under the umbrella of PGMs coupled with the fact that PGMs is an active area of research, one has to prioritize as to what topic one needs to focus on before diving deep in this course.
Beginner level
If you are new to PGMs and have a decent background in probability theory, then I would highly recommend that you go through these](http://web.mit.edu/6.008/www/videos/index.html) videos first. These lectures also are part of the Computational Probability and Inference course on edX. In my opinion, it sets the right motivation for why PGMs are useful and it was quite helpful to go through a few of these videos before officially diving deep in PGMs.
Intermediate level and Beyond
Nevertheless, from a research point of view, I found the following references very helpful:
This course taught by Dr. Sontag provides a rigorous holistic overview of PGMs.
This course taught by Dr. Singla provides a good perspective on learning PGMs. It also has many additional references that are very helpful for applying the theory to many research problems.
Two of the most comprehensive references for PGMs are this and this course taught by the Dr. Eric P. Xing and Dr. Zabaras, respectively. They cover a lot of material in a short span of time so it's important that you have some prior knowledge of machine learning in general. Each topic is covered in extreme detail and the video lectures/slides help understand the material better. In my opinion, it brings all of Machine learning together and after taking the course you realize that most of the specific machine learning models that are covered in standard ML courses are just very specific kinds of probabilistic graphical models. I felt that these two courses help builds the right perspective as an ML practitioner and therefore, the long videos are worth every second of your time.
These course notes by Stanford University are also a very handy reference. Finally, this COURSERA specialization was also very helpful and serves to provide sound theoretical knowledge and also gives the opportunity to apply the inference and learning algorithms to various practical problems. Highly recommend the specialization.