The purpose of this course project is to provide you with a hands-on experience with graphical models and to apply (and possibly, to develop further) the techniques discussed in lecture and homework. For your project, you can choose any topic that would involve graphical models, whether it is an application of graphical models to the domain of your interest, an implementation of a graphical model algorithm (related to whether inference or learning) to evaluate its performance, or a development of a new method related to graphical models. As long as your project idea has a graphical model as an essential part, it is likely it can be turned into a project for this course.
It is preferred that you work on the project in pairs; however, if you cannot find a partner, and if your proposed project is very dissimilar from other proposals, you may be allowed to pursue a solo project. You will have roughly 10 weeks to work on a project, so please plan accordingly. To help you along with a project, we will follow the following (tentative) timeline:
The presentations will be 15+2 minutes format (presentation + 1-2 questions). If two people participate in the project, I expect both to participate in the presentation. As we will be quite short on time, please email me your presentation before class so that I can upload them on the laptop/computer (even 30 minutes before class will do although the night before is safer). For the talks make sure to:
The talk will be evaluated on clarity.
The structure for the final report should be quite similar to that of the talk. The final project will be evaluated on the proposed approach (how much was accomplished taking into account the difficulty of the problem, the difficulty of the implementation, and the novelty of the approach), explanation of the results of the empirical evaluation, and clarity of the written report and oral presentation. As a reminder, the report has to be no more than 9 pages long (with page 9 consisting only of references), and if possible, in the NIPS style format (see projects page). Page limit will be enforced. If two people were working on the proposal, please include a paragraph describing who did what.
Please use NIPS format and style files for your final submission. 8 page (plus one for references) limit will be enforced.
Ideas for possible projects (incomplete, will be expanded over the next several weeks):
Uncertainty in Artificial Intelligence (UAI) is perhaps the most prominent venue devoted in large part to problems related to graphical models. Reading recent papers may give you an idea for a course project.