Probabilistic Graphical Models
Daphne Koller, Professor
In this class, you will learn the basics of the PGM representation and how to construct them, using both human knowledge and machine learning techniques.
Announcements
End of Week 1 and Welcome Community TAs!
Congratulations on hanging on through the first week of class! Starting a new class is always difficult - there are brand new terms to learn, new concepts to get your head around, and for our programming assignment, new software and code bases to grapple with. This will only get better, as we spend the rest of the class fleshing out Bayesian networks and Markov networks; already, some of you have asked on the forums about how to perform inference and learning on PGMs, which we'll spend a good bit of time on in the rest of the class. If you've been busy learning the ropes of Octave and our submission system, you'd also be happy to know that we'll be using the functions you wrote for this assignment in the rest of the class, so the time you took to understand the code and get everything to work was time well spent.
I also wanted to take a moment to thank PGM's 18 Community TAs: John, Miguel, Elena, Misty, Mikhai, Michael, Hans, Michalis, Andrey, Alicja, Anna, Shahar, John, Hari, Ian, Zarutskiy, Willem and Binesh! These students did exceptionally well last time we taught PGM and have been invited back to help make sure you survive as well. I'm sure you've met them already in the forums, and they have already proven to be a great asset to the course. Thank you, Community TAs.
We hope you've enjoyed the first week, and that your appetite has been whetted for what we'll cover in the remaining weeks!
Daphne
I also wanted to take a moment to thank PGM's 18 Community TAs: John, Miguel, Elena, Misty, Mikhai, Michael, Hans, Michalis, Andrey, Alicja, Anna, Shahar, John, Hari, Ian, Zarutskiy, Willem and Binesh! These students did exceptionally well last time we taught PGM and have been invited back to help make sure you survive as well. I'm sure you've met them already in the forums, and they have already proven to be a great asset to the course. Thank you, Community TAs.
We hope you've enjoyed the first week, and that your appetite has been whetted for what we'll cover in the remaining weeks!
Daphne
Sun 30 Sep 2012 3:50:00 PM PDT
Welcome!
We are immensely excited that all of you are joining us, and we look forward to our ten-week-long journey through the land of probabilistic graphical models (PGMs) together. It's going to be a fair bit of work: the Stanford students average 15-20+ hours per week on the class. It'll be challenging, but we'll do our utmost to make sure that you find it fun and rewarding as well.
The goal of this class is twofold: to bring across the foundational ideas and concepts behind probabilistic graphical models, and importantly, to equip you with the tools to use and apply these models to the projects that you care about. To accomplish these goals, we have prepared a rich set of materials, which include video modules that cover many of the key ideas in the field of PGMs, each augmented with quizzes to help reinforce your knowledge.
We have also prepared a series of 9 programming assignments, each focused on a real-world application of probabilistic graphical models, from genetic counseling to recognizing human actions with the Xbox Kinect. By the end of the course (or even sooner!), you'll be ready to bring your knowledge out of the classroom and into the rest of the world. The first programming assignment will be up by the end of the day - in it, you'll be building your own Bayesian network to help a bank solve a credit insurance problem. For most programming assignments, you will need to use either Octave or MATLAB; if you are unfamiliar with Octave, we have provided a set of Octave video tutorials, courtesy of last year's ML-class. Installation instructions for Octave can also be found on the navigation panel to the left.
In addition, we'll have weekly problem sets to help you master the course material; as with the programming assignments, the first set of questions are already up. This week's assignments are all due three Tuesdays from the official start of class, on Tuesday 9 October. This class officially starts on Monday 24 September, and the second week's materials will go out on Friday 28 September, at 8am PDT.
Finally, at the end of the course we'll have a final exam, which will allow you to integrate ideas you've learned throughout the course and test your mastery of the material. The exam will remain open for one week(Nov 30 - Dec 7); you'll be able to log in at any time during that week to take the exam.
After the exam, we'll be giving out statements of accomplishments. We'll calculate the course score in two ways:
More details on the course format, etc., can be found on the Course Logistics page to the left; likewise, we've posted the full course syllabus, and the prerequisites for this course. If you have any questions or feedback, or if you'd just like to say hello, please feel free to use our discussion forums. We also have our own PGM-class wiki at https://share.coursera.org/wiki/index.php/PGM:Main! If you have the time, please contribute to it; you'll not only help out your fellow students, but enhance your own understanding of the material.
No textbook is necessary for the class; we've designed it to be self-contained! However, if you'dd like to read beyond what's covered in this class, you can find a much more comprehensive treatment in the book "Probabilistic Graphical Models", by Koller and Friedman, and published by MIT Press. MIT Press has generously provided a discount code for students enrolled in this course (following a suggestion made by one of the students enrolled in this class).
This class was half a year in the making (building on top of 15 years of teaching this material at Stanford, including three years in an online format), and we sincerely hope that you will enjoy taking it as much as we've enjoyed creating it. We're humbled by the amount of support all of you have shown for this, and we know how valuable your time is, so thank you for spending it with us; we hope that you'll find it worthwhile. And on that note - it's time to start our adventure!
Daphne
P.S. If you're seeing this via email, you'd want to click on this link to the class.
The goal of this class is twofold: to bring across the foundational ideas and concepts behind probabilistic graphical models, and importantly, to equip you with the tools to use and apply these models to the projects that you care about. To accomplish these goals, we have prepared a rich set of materials, which include video modules that cover many of the key ideas in the field of PGMs, each augmented with quizzes to help reinforce your knowledge.
We have also prepared a series of 9 programming assignments, each focused on a real-world application of probabilistic graphical models, from genetic counseling to recognizing human actions with the Xbox Kinect. By the end of the course (or even sooner!), you'll be ready to bring your knowledge out of the classroom and into the rest of the world. The first programming assignment will be up by the end of the day - in it, you'll be building your own Bayesian network to help a bank solve a credit insurance problem. For most programming assignments, you will need to use either Octave or MATLAB; if you are unfamiliar with Octave, we have provided a set of Octave video tutorials, courtesy of last year's ML-class. Installation instructions for Octave can also be found on the navigation panel to the left.
In addition, we'll have weekly problem sets to help you master the course material; as with the programming assignments, the first set of questions are already up. This week's assignments are all due three Tuesdays from the official start of class, on Tuesday 9 October. This class officially starts on Monday 24 September, and the second week's materials will go out on Friday 28 September, at 8am PDT.
Finally, at the end of the course we'll have a final exam, which will allow you to integrate ideas you've learned throughout the course and test your mastery of the material. The exam will remain open for one week(Nov 30 - Dec 7); you'll be able to log in at any time during that week to take the exam.
After the exam, we'll be giving out statements of accomplishments. We'll calculate the course score in two ways:
- Advanced track: programming assignments (worth 63%), problem sets (worth 25%), and final exam (12%). Students who achieve a reasonable fraction of this (probably in the ballpark of 70%) will receive a statement of accomplishment from us, certifying that you successfully completed the advanced track.
- Basic track: problem sets (worth 67.5%), and final exam (22.5%). Students who achieve a reasonable fraction of this (probably in the ballpark of 70%) will receive a statement of accomplishment from us, certifying that you successfully completed the basic track.
More details on the course format, etc., can be found on the Course Logistics page to the left; likewise, we've posted the full course syllabus, and the prerequisites for this course. If you have any questions or feedback, or if you'd just like to say hello, please feel free to use our discussion forums. We also have our own PGM-class wiki at https://share.coursera.org/wiki/index.php/PGM:Main! If you have the time, please contribute to it; you'll not only help out your fellow students, but enhance your own understanding of the material.
No textbook is necessary for the class; we've designed it to be self-contained! However, if you'dd like to read beyond what's covered in this class, you can find a much more comprehensive treatment in the book "Probabilistic Graphical Models", by Koller and Friedman, and published by MIT Press. MIT Press has generously provided a discount code for students enrolled in this course (following a suggestion made by one of the students enrolled in this class).
This class was half a year in the making (building on top of 15 years of teaching this material at Stanford, including three years in an online format), and we sincerely hope that you will enjoy taking it as much as we've enjoyed creating it. We're humbled by the amount of support all of you have shown for this, and we know how valuable your time is, so thank you for spending it with us; we hope that you'll find it worthwhile. And on that note - it's time to start our adventure!
Daphne
P.S. If you're seeing this via email, you'd want to click on this link to the class.
Mon 24 Sep 2012 3:00:00 AM PDT
No comments:
Post a Comment