Please log in to see the topic assignments!

Just click the "My University" button (top right of page – the one with the user icon, second from right) and enter your university credentials. "My University" is not very useful, but once you are logged in, you can see the topic assignments.

Picking Topics

To choose you topic, please send us an email by Wednesday, 2018-04-25, 24:00 listing 3 topics, in your order of preference, along with 2-3 sentences each explaining why you would like to take that particular topic. Please copy the full text of the topic, eg. "Robust Physical-World Attacks on Deep Learning Models", not just "the car topic" because the latter tends to be ambiguous.

Regulations

  • The grade for the seminar is composed of the presentation (60%) and the report (40%).
  • We expect you to be present for almost all talks, unless you have an important reason you cannot attend, such as being ill. The exact regulations are:
    • If important issues are preventing you from attending, please provide some evidence, such as a doctor's certificate if ill. Such absences are completely ignored. This is essentially the same as being ill for an exam.
    • When nothing important prevents you from attending the talks, we expect you to be present. You should not skip more than one day of the seminar. 2 or more absences can lead to a failing grade.
    • According to the department's regulations on exam registration, if you fix a date for a seminar presentation but do not hold the presentation, we have to give you a 5.0. Therefore, if you are ill, please send mail in advance and later bring a doctor's certificate.
  • If you find better papers that are more suitable for the topic, or find the provided papers to be unsuitable, please do contact us to discuss. Information competence, ie. judging how trustworthy a certain publication is, is a very useful skill we would certainly like to promote.
  • Please note the department's guidelines regarding plagiarism. Short version: Plagiarizing something is a failing grade (5.0), and can get you removed from the university (Exmatrikulation). So don't do that.
    • This also applies to slides; please try to properly attribute all images and citations you did not create yourself.
  • The Schreibzentrum can provide you further assistance regarding the writing process (including how to avoid accidental plagiarism). There is also a course on scientific practices for students taught by Barbara Pampel.
  • If you have any other questions, feel free to send us an email or, even better, see us in the office (Z809/10).

Presentations

  • Talks should be about 25 minutes, followed by 15 minutes of discussion. In practice, this adds up to 45 minutes.
  • Talks should be a summary of the most important results of the paper, but also provide enough context for the other participants to understand the topic.
  • Slides must be sent in on Thursday one week before for discussion and feedback. We will discuss these with you after the seminar. If that isn't possible, please make a separate appointment for this discussion, preferably on Friday.
  • Presentations should end with a summary slide containing the most relevant results from the paper. This is more helpful for discussion than the classical ending slides reading "Thank you for your attention" and/or "Any questions?" so please don't use these.
  • Slide sets should use at least 50% slides using graphical or other non-textual elements to deliver their content.
  • Please do not use more than 0 (zero) slides containing the full text of what you are going to say in the presentation (unless you need to show a definition or something similar that you would like to discuss).

Reports

  • The text should be a review of the paper, that is, a short summary of the most important content (approx. ⅔ of the review), followed by your assessment of the paper (approx. ⅓ of the review). Note that you do not have to provide the full context in the report -- you can just cite the relevant literature instead.
  • A draft of the report must be sent in 2 weeks after the presentation for discussion and feedback. Please send email To: matthias.fratz@uni-konstanz.de, marcel.waldvogel@uni-konstanz.de. We will get in touch as soon as we're ready to give feedback.
  • After feedback, the final report has to be sent in 1 week after the feedback discussion.
  • The final report must be 4k-6k characters (not words!), which is 2-4 pages depending on font size and margins. Please try to make this text short and to the point. Send email To: matthias.fratz@uni-konstanz.de, marcel.waldvogel@uni-konstanz.de.
  • If you include an abstract, table of content etc., they do not count against the character limit.
  • Graphical illustrations are welcome and count as a single character each.
  • The report must be in English and in PDF format. The rest of the format is free, though single-spaced at a readable font size is preferred.
  • References must be present and must include at least author, title and year of publication. Scientific articles should also include the publisher, journal name, volume / issue, pages etc. Web references obviously require a URL and when it was accessed. Using bibtex is recommended, because it reduces the likelihood of missing fields.

Timing Overview

t-8 days

Slides sent to us (please make sure concepts are clearly presented; this is more important than details of the algorithm etc.)

t-7 days

Slide discussion & feedback (after seminar, if possible)

t±0 days

Presentation

  • 20 - 25 minutes presentation, including personal opinion
  • 15 minutes discussion (good questions from the audience will result in a bonus for the asker)
t+14 days

Draft report sent to us (4000…6000 characters; including opinion)

t+k days

Feedback by us

t+k+7 days

Final report due

How to get good grades

  • Make sure you understand the contents of the paper well enough to explain the key points in your own words.
  • Focus on the most important parts of the paper. In 20 minutes, you will not be able to present everything. If you are unsure whether it is ok to focus on some particular aspects of the papers, just ask us.
  • Present the topic in an understandable way. This may involve simplifying things, which is fine as long as you mention it.
  • Structure both your presentation and your report so they follow a common theme. This massively helps understanding.
  • Look for related literature. Some topics explicitly mention that you have to find related papers, however all topics profit from some overview of the research area.
  • If you have references that you read to get some knowledge about the research area, please feel free to cite those as well. You would not normally cite such literature in either a paper or a thesis, however they are very helpful in a seminar report.

Criteria for a Good Presentation

In no particular order:

  • Presentation Skills: Talk is delivered in understandable English and with suitable slides. Speaker speaks with confidence and it is easy to follow.
  • Time Management: Talk is adapted to the available time and thus neither hectic nor excessively wordy.
  • Structure and Understandability of Content: Talk follows a common theme, is well structured and supported by slides that help understanding. The content is easy to understand.
  • Own Thinking: Talk is a structured summary of the topic that has been well throught-out. The interrelations between the parts have been considered.
  • Scientific Correctness: Own thoughts are coherent and substantiated; facts and statements are reproduced correctly.
  • Focus: Aspects of the topic are treated according to their relevancy for the overall topic. No parts necessary for understanding are omitted. No unnecessary details.
  • Kolloquium: Presenter has some subject knowledge beyond the content that was presented and can answer questions in an understandable way. (Note: This related to reasonably on-topic questions only!)

Criteria for a Good Report

In no particular order, but sometimes suspiciously similar to the criteria for a good the presentation:

  • Structure and Understandability of Content: Text follows a common theme, is well structured and conveys the content well.
  • Information Competence: Literature list contains relevant literature and follows the proper conventions for scientific citations. No irrelevant literature. Sources are properly quoted, both when quoting literally and when reproducing content.
  • Own Thinking: Text is a structured summary of the topic that has been well throught-out. The interrelations between the parts have been considered.
  • Scientific Correctness: Own thoughts are coherent and substantiated; facts and statements are correctly reproduced.
  • Focus: Aspects of the topic are treated according to their relevancy for the overall topic. No parts necessary for understanding are omitted. No unnecessary details.
  • Linguistic Adequacy: Language is suitable for a well-written scientific text. No use of narrative or newspaper style, or excessive use of colloquial language.

Materials from the Writing Center and beyond

Some advice on how to do good presentations:

  • Video: How NOT to give a presentation. Research Skills class at the University of Cambridge, 2012, by Neil Dodgson. You may want to take a look at the related videos as well.
  • Avoid long text or lots of bullet points. Prefer pictures over text.
  • Use colors for linking where possible, ie. same color for related items.
  • Only use animation or other effects if it helps understanding.
  • Rehearse your talk sufficiently to iron out any problems and to make sure you are on time.
  • Use pictures to explain the content if possible.