Stay Here for Tips and Instructions
Student-led presentations start 2/12, first batch runs 2/12 through 3/25.
By January 31, 5 pm, sign up for a presentation slot (from below, on the dates 2/12 through 3/25 only): Form groups of 2, or discuss with Indy at one of his office hours (Tu, Th 10.45 am-12 pm, 3112 SC). To sign up for a presentation slot, you have to meet Indy during one of his office hours (email will not be accepted). See the table below for a list of available topics. For your Topic of interest, if the "Presenters, Slides and Reviews" link for that topic's row (see table below) is empty or has a "& ??" on it, that topic is still available!
O Tips for Presentations and Reviews
These tips are outlines only, so that you have maximum flexibility and room to improvise!
Tips for Preparing a Presentation (Do improvise!): Typically, 15 slides per paper. 2-3 slides on motivation and background. 5-7 slides on core ideas of the paper. 3-5 slides on experimental data. 2-3 slides on your thoughts/criticisms/questions/discussion points about the paper.
Use examples - the more and the earlier, the better. It improves clarity.
Use pictures - the more, the better. It improves clarity. Use animation, but not too much.
Write concrete bullet points that make sense and are to the point - don't be vague or general.
Remind the audience of definitions of basic terms before you use them in your presentation. Define before use.
It is ok to reuse figures from the original paper/presentation by the authors of a paper. It is NOT ok to reproduce entire slides or slide bullets - the presentation outline and execution should be original. Do feel free to reproduce ideas and questions (e.g., from the student reviews) though, as long as these are in your own words.
When presenting experimental plots, make sure you state what the experiment was exactly, and what the parameter values were (best if on the slide itself). An experimental plot is meaningless without a description of the experiment!
When presenting your questions/criticisms/discussion points, be as constructive as possible - respect the paper but also your own opinions.
Use text only where absolutely necessary - it puts folks to sleep and makes them ask more nasty questions.
Jump into the "meat" of the paper as quickly as possible - within the first 3 slides is best.
Anticipate questions that you're likely to get - better still, prepare slides that have answers (or non-answers) to these questions.
If the paper being presented is not yours, try to do justice to the authors' contribution, but also make sure you state alongside what your personal/research opinion is about each of the authors' design decisions (this can pre-empt questions, esp. if the design decision is questionable).
Keep track of time, because it is the only thing guaranteed to increase monotonically.
Tips for Writing a Review (Do improvise!): 2 pages max (total). For each paper, 1 paragraph on the core idea of the paper, followed by list of pros and cons of the approach, your thoughts on how the paper can be further developed, and any questions/criticisms/thoughts/doubts/wanderings about the paper.
DO NOT comment on the English/Grammar/typos in the paper. We are interested only in criticism of the material, problem, ideas, algorithm, approach, experiments, and methodology in the paper. Please constructively criticize the work not the diction. After all, the paper has been accepted, and we're trying to understand why it was (or why it shouldn't have been) technically.
It is ok to comment on experiments the authors ought to have done.
O What to Do for Each Class (February 12 onwards)
Instructions for presentation sessions (discussed in class):
Each Group of Presenters: Based on the "Main Papers" for your session, prepare a 45 minute presentation and a short list of topics for class discussion. The presentation should cover ALL the "Main Papers" marked for that session. Meet Indranil by 5 pm the day prior to your presentation to show slides - please email for an appointment first!
Papers in each session are related by a few common threads. One thread is evident from the topic name of the session. See if you can discover the other common threads as you read along.
Since the papers in each session are related, they will take a short time to present together. Your presentation need not cover everything in every paper, but neither should it be superficial.
Presenters need not write reviews for their session.
Others (non-presenters): Write a 1 to 2 page review of any TWO of the "Main Papers" for that session. Your review should be at most 2 pages total, not per paper. Review should contain brief summary/findings from the "Main Papers" for the session, your comments and criticisms (try to be professional while writing the latter), and any ideas for further work. Email indy at ad dawt uiuc dawt edu with your review in the message body (ascii text only! no attachments!) with the subject line "525 review mm/dd" by the start of class. Emails not adhering to this standard may not be considered as submissions (think: automatic scripts). Also turn in a hardcopy of your review during class. Reviews are written individually and not in groups.
Although reviews will be graded, their main purpose is to start you thinking about the paper(s) critically and creatively. As you read along, it might help to write footnotes about new principles and design techniques that you are learning.
Note: Some papers may be accessible from only inside the CS department.
Your work (presentations and reviews) should be original and independent. If you reuse slides from someone else's presentation, please acknowledge them. The standard university policies on original work, cheating and attribution apply to these presentations.
A final note to all: You are encouraged to read ahead of the class, especially if you find one (or more) topics that interest you enough to want to do a project and write a paper on. As a tip, some of the more interesting topics occur in the second half of the course.
Report broken links to indy at cs dawt uiuc dawt edu
(c)ig