Seminar Advanced Multimedia Systems (2010-11, term 3): Home

TERM 3, 2010/2011
INFOAMS: Seminar Advanced Multimedia Systems
Timeslot D, 7.5 ECTS

Lecturer: Wolfgang Hürst (WWW, Email)

Home · Procedure & grading · Schedule

General procedure

Besides learning the content, the purpose of a seminar is to gain and demonstrate skills such as how to look into, read, and understand scientific literature and how to present, evaluate, and discuss scientific work. Consequently, you will have to read research papers and give a presentation (covering one paper) in the meetings. Final grades will be based mainly on the presentations but other issues will influence them as well. Due to the special circumstances this year, participants of the seminar also have to take the first part of the exam of the INFOB1OIS course which will count 50% for their final grade (cf. related comments below).

After a general introduction from the instructor in the first meetings, individual topics will be assigned. This year, each participant has to pick one paper from the following conferences:

MobileHCI (years 2009 and 2010)
ISMAR (years 2008, 2009, and 2010)

After that, you have some time to prepare your presentation using the selected paper and related literature (which has to be researched by yourself). In addition to the presentation, you have to write a short review for the paper of your presentation, and one review for each of the mandatory papers handed out at the beginning of the seminar. Presence in the meetings is mandatory and it is expected that you participate in the discussions after the presentations.

Please observe the deadlines specified below as missing them can result in a reduction of your final grade. Also cf. the general grading rules at the bottom of this page.

Presentations

Your presentation should be based on the paper that you selected from the conferences listed above. If necessary, you are welcome to include other papers (e.g. if you want to go deeper into a specific aspect) as long as it is related (i.e. the talk should cover one single topic). Everyone has to prepare slides and give one presentation. Each presentation is about 30 minutes long (+/- 5 min, keep the time limit!) and followed by a discussion in which the audience can ask questions and the papers are discussed. It is the task of the presenter to moderate this discussion and encourage questions from the audience.

Grading is mainly based on the content of the presentation. However, presentation style, presentation material, etc. influence the final grading as well (weighting content vs. style approximately 3:1). In preparing your talks, use a format derived from your understanding of the topic. Do not copy every tenth line of the article, but present for example the statement of the problem, alternative solutions, evaluation of alternatives, elaboration of chosen alternative, results, discussion, pros and cons, etc. The bottom line is that it must be clear that you understand what it is about. Be precise. Do not use phrases like "the method works reasonably well". Note that you are encouraged to be critical, e.g. by discussing things such as: Are the used evaluation methods convincing? Do the results generalize? Etc. But focus on the content, not style etc. (criticizing that should be done in the reviews).

The following hints might be helpful in preparation. However, I highly recommend not to accept anything without criticism just because "the sources say so". In my opinion, such advise should only be seen as a general guideline. In fact, all of the references given below contain some arguments which I personally disagree with. Nevertheless, it might be useful to have a look at them and other, related resource.

Simon Peyton Jones, John Launchbury, John Hughes: How to give a good research talk (HTML, PDF.gz)
Frank R. Kschischang: Giving a Talk - Guidelines for the Preparation and Presentation of Technical Seminars (PDF)
Tia Newhall: Some tips for preparing a research presentation (HTML)

You can also look around on the web for yourself. There are hundreds of documents containing useful advice on how to give a good presentation (but there's also lots of junk out there and rather questionable suggestions, so again: be critical and don't take anything for granted just because it is written somewhere). Here are some rather amusing examples on how not to do it:

Deliverables and Deadlines: Slides should be send to the participants via email on the day before the presentation. You should add a list of papers that you read in preparation of your talk.
Influence on grading: approx. 35%

Reviews

Although scientific literature usually undergoes a review process, the quality of papers and articles can vary significantly. In addition, scientific results are normally presented in a certain context. Therefore, it is always important to have a critical look at the presented work and challenge the results (e.g. do the results generalize or depend on a particular context, why didn't the authors mention anything on time complexity of their proposed approach, etc.). In addition, there is almost always room for improvement. Thinking about what could be done to improve a paper (e.g. can the graphics be improved, would the paper benefit from a different structure, etc.) can also help you later when you have to present you own work sometime (e.g. your master thesis).

In order to train your ability to evaluate, assess, and judge scientific literature, participants have to write one review for the paper they are presenting and the two papers for mandatory reading that will be handed out in the first meeting. Use the review form that was distributed via email.

For evaluating the quality of a paper, it might help to have a look at the following references which give hints on how a "perfect" paper should look like:

Roy Levin and David D. Redell: How (and How Not) to Write a Good Systems Paper
Simon Peyton Jones: How to write a good research paper (slides in PDF)

Similar to the references for giving good presentations indicated above, these should only be seen as general guidelines and exceptions exist.

Deliverables and Deadlines: The reviews of the mandatory literature should be send to the instructor via email on March 8, 2011 (i.e. one day before the 2nd meeting). The reviews of the paper used for your presentation should be send to the instructor via email the Monday before your presentation (i.e. on Mon, 28.3. for people presenting on Fri, 1.4, and on Mon, 4.4. for people presenting on Fri, 8.4.).
Influence on grading: approx. 10%

Participation in the meetings

Attendance in the meetings is mandatory. In order to guarantee audience participation in the discussions after the talks, it is required to ask questions. The quality of the questions will influence the final grading. For this, everyone has to write down at least two questions per talk. These questions are collected at the end and rated. There won't be enough time for everyone to ask a question after every presentation. However, you should not only write down, but also ask questions every now and then.

The goal of the questions after the presentations is to clarify open issues and problems, etc., but also to discuss the paper (is the approach valid, how good does it work, is the problem relevant at all, etc.). Normally, the discussion starts with questions about the content and naturally evolves into a more general debate. Hence, the presenter should not only be able to answer questions, but also become more of a moderator as the dicussion evolves. In order to do this, you could prepare some general statements and issues to foster the discussions.

Deliverables and Deadlines: Questions are collected at the end of each meeting.
Influence on grading: approx. 5%

INFOB1OIS exam

Because of the special situation this year, participants of the seminar have to take the part of the INFOB1OIS exam that covers the first five lectures (i.e. the ones given by W. Hürst). Note that you are also entitled to participate in the related experiments of the INFOB1OIS course. This can result in a bonus of your grade of up to 0.5 (cf. related info on the INFOB1OIS website).

Deliverables and Deadlines: The exam will take place on April 20, 2011, 17-20h, in EDUC-GAMMA.
Influence on grading: exactly 50%

Exam / grading:

Grading will depend on the exam (50%), the quality of the given presentations (35%), the quality of the written reviews (10%), as well as the participation in the meetings (5%). Please note that the weighting related to the presentation, reviews, and participation is only a rough guideline and that we reserve the right to (slightly) modify it in case you handle one part particularly well or bad. The related bonus from the OIS experiments will be added before rounding (if applicable).

2nd exam / retake:

If you fail in the INFOB1OIS exam (i.e. your grade is less than 5.5), you can participate in the related retake if your grade is at least 4.0. In case of an insufficient performance considering the issues related to the presentation (i.e. the actual presentation, the reviews, and the active participation), 2nd chances will be handled individually. For example, if a presentation is not good enough in order to pass the course, students might make up for this by giving another one. It is up to the instructor to specify the amount and form of the individual effort required for such a 2nd chance. You have to do "reasonably well" in order to qualify for a 2nd chance, e.g. by having at least a grading of 4.0 in your 1st trial.