PhD Defense Slides and Lessons Learned

In July this year I finally defended my PhD which mainly focused on (adversarial) robustness and uncertainty estimation in deep learning. In my case, the defense consisted of a (public) 30 minute talk about my work, followed by questions from the thesis committee and audience. In this article, I want to share the slides and some lessons learned in preparing for my defense.

My PhD defense marks the successdful end of a four year journey in machine learning and computer vision research. While defense vary widely across institutions and universities, I had to prepare a 30 minute talk that is followed by questions from the committee and audience. In addition to the difficulty of squeezing several papers into a 30 minute polished talk, there was also a significant delay between submitting and defending my PhD — roughly 4 months. While I can't share the final thesis yet, you can find my defense slides below, followed by some lessons learned in preparing for my defense.


The slides include some GIFs, so with 27MB the download might take a second. Also, make sure to view them in a PDF viewer supporting animations:


Lessons Learned

Preparing for my defense took significantly longer than I initially estimated. I thought it could not be too hard to put together slides of all my projects, given that I already gave a few talks and had slides for most of my papers ready. However, preparing a polished talk that highlights achievements, giving a good high-level intuition while at the same time demonstrating technical knowledge, was incredibly difficult and took multiple iterations. Besides, I wanted to talk to steer the committee's questions to some extent and make sure I will stick to the 30 minute limit. Here are some lessons I learned surrounding the preparation:

  • Start preparing early enough: Even though I thought the preparation to take less time, I took some long weekends roughly one and a half months before the defense to start preparing the slides. Looking back, this was a great decision as I ended up working on my slides until the last couple of days before the defense.
  • Schedule a practice talk: In my case, a practice talk really made a huge difference. Initially, I included too much content in my slides and it was good to get objective feedback on what to focus on, where more details and time is required and which topics can be covered with less details. I had the practice talk two weeks before the defense and I really used these two weeks to improve my presentation.
  • Practice: Before the practice talk, I practiced the talk and timed it to the minute. This really came in handy in the end because it reduced stress as I knew exactly how to start or get back into the talk on each slide. During the talk this was also helpful as interruptions didn't break the flow of my presentation.
  • Backup slides: Even though I did not end up using most of my backup slides, preparing them helped me to think about potential questions that the committee might ask.
  • Test the setup: A day before the defense, I tested the technical setup. As my defense was hybrid, there was a lot that could go wrong — the virtual meeting room, the microphone etc. It really reduced stress to have it tested once before the defense. Some things that are important to keep in mind is that the setup should also take into account discussion and questions of the committee/audience which are sometimes difficult to handle in a hybrid setting.

Regarding the content of the talk, I found the following to work well — and I found a similar scheme can be found in many other defenses, as well:

  • Fuse motivation with the outline: I ended up putting a lot of effort into a good motivation of my talk that also happened to reflect the outline of my talk. This really helped to tell a consistent high-level story throughout the talk.
  • Highligh achievements: I feel that it was important to highlight the papers that I have published over the years — even those not included in the talk.
  • Technical detail: For two projects, I decided to go into technical detail. While I did not have many technical questions, I got the impression that this was expected by the commitee. Of course, it was tricky to decide which projects to highlight in detail as this usually means that all other projects can only be covered on a higher level.
  • Ignore some projects: In order to go into technical detail for few projects, I decided to not talk about three of my papers. I just mentioned this line of work in my conclusion and the introduction. Initially, I found it difficult to let go of these parts, but it helped to improve the focus of the talk.
  • Have some slides on future work: An easy way to steer the discussion/questions after the defense talk are future work topics. Also, I found that these slides can be used to end the talk on a positive note by expressing your excitement about future research — note that I had these slides after the conclusion.
What is your opinion on this article? Let me know your thoughts on Twitter @davidstutz92 or LinkedIn in/davidstutz92.