"Assigned" vs. "external" papers:
In the following, an "assigned" paper is the paper assigned for the entire class to read and review. The proponent/opponent discussion leaders and the experiment presentations will focus on these assigned papers during in-class discussion. An "external" paper is a selected paper on the syllabus that is presented by a student presenter that day; external papers are optional reading for everyone else in the class.
Notes about advance
Please coordinate in advance with the other student presenters on your day to ensure that no single paper receives two experiments, that no single paper has more than one proponent/opponent, and that no external paper is covered twice. You can find contact info in the UT directory, or ask the instructor if you have any problems.
Note that draft presentations (for both experiments and external papers) are due **one week before** the slot your presentation is scheduled. This is a real deadline. This means you will need to read the papers, prepare experiments and/or slides, etc. more than one week before the date you are signed up for. The idea is for the instructor to preview a draft ahead of time, so that we can iterate as needed the week leading up to your presentation. Having some placeholders is fine, but the submitted drafts should clearly illustrate your plans and scope, and they should include visuals/figures or mockups indicating what will be there.
To avoid any surprises or delays, it's always wise to check your slides on the projector in our seminar room before class starts.
reviews and discussion points
The quality of our discussions will rely on how prepared everyone is when they come to class. It is important to do the reading in order to actively participate. Students are required to submit two paper reviews per week (and their associated discussion points) for the assigned papers. We'll usually have 2-3 assigned papers each week; choose 2 of those assigned to review. In weeks that you are presenting an external paper or an experiment, you can skip writing the reviews. Note that this does not apply to weeks where you are leading the for/against discussion; proponents/opponents should still write reviews as usual.
Each paper review should address most of following (in any order):
- Give a summary of the paper in your own words (very brief, 2-3 sentences)
- What is the main contribution of the paper?
- What are the primary strengths and weaknesses of the paper?
- What is inspiring? (about the problem statement, the approach development, etc.)
- How convincing are the experiments? If something specific is lacking, what should have been tested?
- Describe one specific way in which the work could be extended.
- Additional comments, including unclear points, connections you see between the papers
Keep in mind it's often easier to see flaws in a piece of work in hindsight, or judging from a distance -- yet we strive to comment on the good and creative things, too.
After writing your reviews, extract 1-2 favorite "discussion points" based on your analysis. These should be key observations, lingering questions, interesting connections, etc. that fall out of your full paper review. Think of it as a recap of the most salient aspects of your paper review. Typically this part should be about 2-4 sentences for each of the 2 papers you reviewed.
Reviews and discussion points are due by Monday at 8 PM. Instructions for submitting reviews on Piazza are here.
Prior to class, we encourage you to browse your classmates' discussion point postings and (before or after class) share reactions on Piazza.
Each assigned paper will have a designed proponent and opponent. As proponent/opponent you are responsible for initiating discussion about the paper, with an emphasis on strengths/weaknesses of the work, respectively. Come to class prepared with 5 points of discussion (in favor or against the paper). You may draw on the discussion points posted by the class on Piazza. Please do not bring slides. In class the discussion is dynamic; almost certainly you will not sit down and enumerate your list of prepared notes. Rather, you will help initiate our discussion and inject items at the appropriate moments---and only if they have not already come up.
Please coordinate in advance with other proponents/opponents to ensure no single assigned paper has more than one opponent or proponent.
Each student will give a presentation in class covering one of the designated "external" papers for the day's topic (marked with ¤ on the syllabus list). Note that this paper will be one that is optional for the rest of the class to read, so the presentation aims to introduce the class to the work. The talk should be well-organized and polished, about 15 minutes. Please run through it beforehand and check the time (a good rule of thumb: generally 15 minutes ~ target max 15 slides total).
Include these components in the presentation:
- Clear statement of the problem
- Why the problem is interesting, important, difficult
- Key technical ideas, how they work, main contributions, strengths and weaknesses
- Evaluation, summary of key experiments and data
- Open issues raised in the paper, likely extensions
- Relate this work to the assigned reading, where relevant
- Highlight example qualitative (image/video) results when relevant
Try to use applications to motivate the work when possible, and look for visual elements (images, videos) to put in the presentation. Check out the webpages linked on the class webpage, and also look at authors’ webpages for supplementary materials like videos or demos. It’s perfectly fine to re-use authors' slides from conference talks etc. when available, but be sure to clearly cite the source on *each* slide that is not your own (e.g., write "Slide credit: Jane Jones" at the bottom of each borrowed slide). Furthermore, even if you are borrowing slides, be aware you may need to edit them to make sure the flow and length are suitable for our class.
Note that draft external paper presentations are due to the instructor **one week before** the slot your presentation is scheduled.
Please bring the slides on your own laptop to class, and then email the final pdf of the slides to the instructor after the class session when you present. Name the file <lastname>_paper.pdf. To avoid any surprises or delays, it's always wise to check your slides on the projector in our seminar room before class starts.
Please coordinate in advance with the other paper presenter to ensure that no single external paper receives two presentations.
For each assigned paper (marked with « on the syllabus list) one person will present the results of some experimental evaluation of some main idea in a paper we read. Basically the goal is to implement a distilled version of an essential technical idea in the paper, and show us some toy example of how this works in practice. For many papers, you may be able to find code or binaries provided by the authors online (see links on our course page schedule alongside the papers). The goal is to help us gain a more complete intuition about the work we are studying. You might:
- Show a simplified example that highlights an expected strength or weakness of the approach
- Experiment with different types of datasets
- Perform an ablation study to understand the method's behavior more deeply
- Compare some aspect between two assigned papers
Note that the goal here is not to recreate published results or to build systems as described in the paper. Instead, you are looking to make a small illustrative experiment that will let us more deeply understand what we have read. It is an experiment, not a demo (though if the authors provide an online demo, do also briefly display this). This presentation should not spend time explaining the methods of the papers. The slides should be about your experiment, not a recap of the paper.
Spend some time playing with your implementation, and put thought into what would be an instructive toy example to show the class. The experiment should allow us to learn something about the method, not just see it. If you needed to implement something yourself, explain how you did it, and especially point out any details or choices that weren’t straightforward, in case others in the class can leverage your experience later when working on the project. Be sure to explain the rationale for the outcomes, and conclude with a summary of the message(s) your example illustrates. In your presentation slides, include links to any existing code, data, etc. you used.
An experiment presentation should take about 20 minutes. Please run through it beforehand and check the time (a good rule of thumb: generally 20 minutes ~ target max 20 slides total).
Note that draft experiment presentations are due to the instructor **one week before** the slot your presentation is scheduled.
Please bring the slides on your own laptop to class, and then email the final pdf of the slides to the instructor after the class session when you present. Name the file <lastname>_expt.pdf. To avoid any surprises or delays, it's always wise to check your slides on the projector in our seminar room before class starts.
Please coordinate in advance with the other experiment presenter to ensure that no single assigned paper receives two experiments.
There are two programming assignments due during the first half of the course. See important dates on our homepage for all deadlines. You will have two weeks to complete each one. All assignments are due via email to the TA by 11:59 pm on their due date. By default, an assignment loses 10 points of credit automatically for every day late.
A project could be built around any of the following, and must be done with a partner:
an extension to a technique we study in class
an in-depth analysis and empirical evaluation of one or two existing related techniques
design a novel approach and perform accompanying experiments
Project proposals will be due before the middle of the term. Details on the format for the proposal and all aspects of the final project (paper writeup, poster for final presentation, video) will be posted online prior to the proposal deadline.See important dates on our homepage for all deadlines.
- 25% participation (includes attendance, in-class discussions, paper reviews)
- 15% coding assignments
- 35% presentations (includes drafts submitted one week prior, and in-class presentation)
- 25% final project (includes proposal, presentation, final paper)