CS 795: Paper Presentation Guidelines

You will have 60 minutes for the presentation; we will reserve an additional 15 minutes for Questions/Answers and Discussion. Understanding and evaluating a research paper is a non-trivial task and it often requires *multiple* rounds of careful reading (and thinking) before you can grasp the main ideas and techniques. As the speaker, you have to be familiar with all the details of the paper in order to prepare an effective presentation (and answer questions), but please keep in mind that the audience has only some basic background in low-power computing (through CS 795 lectures). Thus, introducing the research topic/problem is as important as the technical content.

• Introduction (approximately 20 minutes): First you need to explain the settings and goals of the research. Remember that your audience depends on *you* to understand this research effort, and if you fail to describe the settings/objectives appropriately, then the solutions and evaluations you present will have little value. To make the paper more accessible, you way want to connect the area to the concepts/topics we discussed during the lectures (if applicable), and *also* present some background knowledge that you consider important to follow the paper. As an example, if you present a paper on game-theory-based energy management, then you will need to introduce the fundamental concepts of game theory (perhaps by also adding some helpful examples). After a general introduction to the area, you will need to describe the *specific* problem the authors address. Each research paper will have certain *contribution* claims in the area, and often, this will be explicitly stated in the beginning of the paper. Most researchers do also *position* their paper with respect to other, existing research studies and underline their specific contributions. So, you should also briefly summarize the connection of this research to other papers in the area, as stated by the authors. In many cases, you will need to skim at least a few papers most related to the one your are presenting.

• Technical Content (approximately 30 minutes): This is the main part where you will explain the solutions developed in the paper. Important analytical results (e.g. theorems) and/or empirical evaluations (e.g. simulation results and their interpretations) should be conveyed to the audience with clarity. You do not have time to cover the details of the proofs and mathematical derivations (but obviously you must study *all* the parts of the paper thoroughly to be able to answer questions/doubts, and more importantly, to explain the main ideas). One common mistake is to "cut-and-paste" the main theorems and/or results of the simulations without any explanations. This must be avoided; remember that your task is to make the paper intelligible to the audience. If you prefer, you can copy some results (e.g. important theorems) of the paper, but only when you believe that the audience is ready to follow the paper-specific terminology. Alternatively, you can re-phrase the results obtained by the authors using your own words. In either case, you have the responsibility of helping the audience to "make sense" of the paper through your own explanations.

• Conclusion and Critique (approximately 10 minutes): Having presented the contents of the paper, you will offer your *own* evaluation of the paper. What are the merits of the paper? What are its drawbacks? Is it technically correct? Any problems in the mathematical derivations/proofs? Is experimental evaluation sound, for example, do the authors cover a sufficiently broad (statistically significant) spectrum in the simulations? Is the solution developed in the paper efficient? What do you think about the *value* of this contribution? (Major or minor/incremental contribution?) Can you propose any extensions to this research?

• Discussion and Questions/Answers (15 minutes): This is the phase where the audience will ask questions about the paper (technical content and/or your evaluation). After the speaker tries to clarify any doubts, we (as the entire class) will try to reach a verdict about the paper's contributions, hopefully through a discussion.

Please e-mail your presentation to the instructor by 10 AM on your talk day, so that it can be uploaded to the class computer in a timely manner. 2-3 minutes for a slide is a reasonable estimate, but you can prepare extra slides for potential questions (if you believe that that will help).