We have in our lab these (August 2010) days two thesis proposals and one defense. Here are some tips for these presentations and the associated documents (and also for future ones). Thanks to all the SMLV group for contributing to this list. I will try to keep updating this post in the future and refer future students to it as a central tip repository the SMLV group.

Assembling a Committee and Scheduling

  1. Find out the rules regarding the composition of your committee. In particular, what is the policy for external members (out of your school and out of Georgia Tech).
  2. Get advice about different committee members from your advisor, other students, and the web. The ideal committee member should be in your research area (more or less), friendly (you don’t have to be their friend but you should develop a good working relationship with them), and have time to give you feedback when you need it. They should also have time to attend your presentation (see below).
  3. Scheduling the presentation can be very frustrating. Each committee member has different time constraints and sometimes they don’t communicate them efficiently by email. I suggest first asking for their travel dates. Then identify a week or two and use a tool like doodle.com (where different people can enter their time constraints) to identify a specific date and time.
  4. Start scheduling the presentation date (see above) several months in advance as it may take a while.
  5. After scheduling, email reminders to the committee members to make sure they remember.

Presentations

  1. Remind the committee members the date/time/place the day before the presentation. Make sure you are familiar with the technical equipment (phone conference, projector, etc.).
  2. Speak loudly and clearly and look at the audience frequently.
  3. Aim to speak for no more than 45 minutes assuming you aren’t slowed down by questions. You will probably get questions during the exam in which case try to keep the presentation to less than one hour.
  4. Stand close to the screen and point at the screen occasionally. Use this to help explaining complicated slides where you refer to different parts of it. I prefer this over a laser pointer but the latter is useful for large rooms/screens.
  5. Avoid long and complicated sentences in your slides. Use short ones or sentence fragments.
  6. Bulleted lists are very common in slides. The above tip is important to follow (use sentence fragments instead of long sentences). Also, make the different bullets parallel from grammatical and semantical perspectives (they should use the same tense/person/singular-plural and should voice exchangeable issues). If you bulleted list with more than one level, keep parallelism across levels, that is all bullets in level 1 should be parallel, and the same with level 2 and 3.
  7. Have a slide at the very beginning with an outline of the presentation and division of your thesis work to distinct parts or components.
  8. Have a slide beginning each of the major parts in your thesis including a one sentence summary of your contribution, paper(s) in which it appeared, and collaborators.
  9. Avoid copying propositions and definitions from papers or from the proposal/dissertation document. These tend to be long and wordy. Consider simplifying them for the slides: replace specific conditions with high level descriptions (e.g., sufficiently smooth or nicely behaved instead of precise differentiability and measurablity conditions; you can always elaborate if asked). Also, modify long sentences to sentence fragments.
  10. After the beginning slide (above), have a slide with a problem or an example that motivates your contribution.
  11. Have a conclusion/contribution slide at the end of each of the thesis parts with summary (more detail than the beginning slide).
  12. Make sure the audience understands what is your contribution and what is general outline/discussion/introduction. Also make sure the audience knows what are the major contributions and what are the minor ones. Consider highlighting your contributions and major contributions with different color or format or some other way. If some of your results were hard to get make sure the committee appreciates the difficulties.
  13. Your primary audience here is the thesis committee so you can assume some basic ML knowledge. Minimize motivation/background/setup and just straight to your contribution.
  14. Include brief discussion of future work for proposals (1-2 slides are sufficient). No need to do that in defenses.
  15. Clear the event with the graduate office and make sure you have the necessary form with you (the one the committee members sign).
  16. The committee may ask questions that challenge your work or that are hard to answer. The important thing is to keep your composure and do the best you can. If you don’t understand a question try to get to the bottom of it rather than answering a different question. Answering hard unexpected questions is probably the hardest part of a thesis proposal or defense.

Documents

  1. Make the document double space or 1.5 space. I am not a fan of this personally but this seems to be the common practice for these documents.
  2. Include in the beginning text describing what your collaborators did (no need to include advisor here since work you did together is part of the mentoring process). If you don’t do this people may complain that you unfairly present as yours work that others did.
  3. Include a section at the beginning with a list of your referred publications. Highlight the ones that are relevant to the thesis.
  4. Post the document online and email the committee members a link to it at least one week before the presentation (preferably two). You can keep updating the online document with the latest version. This will ensure that the committee members have access to your latest version without cluttering their emails.
  5. If you want to include an acknowledgements section, wait until after the thesis is approved.