I spent almost half a month preparing, and a lot of that time went into reading the paper Fast exact shortest-path distance queries on large networks by pruned landmark labeling | Proceedings of the 2013 ACM SIGMOD International Conference on Management of Data. I also spent a lot of time making slides, then revised them several times. The slides are here. But when I presented today, it was a complete mess. I was probably the worst of the four presenters.
Being embarrassed is one thing, but the result is genuinely worth reflecting on. It was another one of those “I clearly put in so much effort, but…” moments.
Objectively speaking, I had already sensed the problems before going on stage:
- How could I possibly finish 70 pages of slides in 15 minutes?
- I could not even present my own slides fluently without looking back at them to remember how a proof worked, let alone do it under the nervousness of being on stage.
The final result was not only TLE, but also WA.
For the first problem, I later asked the teacher for advice. The conclusion was that in such a short 15-minute presentation, it is obviously wrong to try to show the audience all the details and fragments of the algorithm. That approach not only buries the key point, meaning the real highlight of the work, but also makes the audience lose interest.
A reasonable presentation should therefore use a concrete example, such as a specific network, to show the audience how the algorithm runs. It does not need to include many rigorous proofs. Instead, it should give readers the intuition that “this is correct.” The details can be left for them to read or ask about later. Within a short time slot, the content should focus on the essential creation and the novel highlight of the work, and it can compare with previous work to make that highlight stand out.
This may look like it reduces the burden on the presenter, but in fact it raises the bar. You need to grasp the core and the shine of the work before you can deliver the right content to the audience. Thinking about it this way, my preparation was still not sufficient. If someone asked me, “What do you think is the highlight of this paper?” I might not be able to answer. That reflects a lack of my own insight into the work.
In that sense, my presentation was almost a complete failure: a failure in direction. That is what made me quite frustrated. My slides were basically a complete reconstruction of the paper, with every formal proof I could provide included. But that is not a good set of slides for a 15-minute presentation. At least in form, it was wrong.
I am grateful that Professor Cheng Gong gave us this opportunity to practice. Even after I presented like that, he still said it was brave and not bad. But this was also my first time giving this kind of algorithm-heavy paper presentation in front of more than a hundred classmates and teachers. It was a major challenge for me, even though I failed the challenge. It was my first time speaking in this setting, and I was indeed a little nervous. These are things that need to be trained again and again.
I think something Professor Cheng said was very right: the training this presentation gave me is tiny in the scale of a whole life. But if a person “perfectly misses” every single one of these tiny opportunities and exercises, the eventual gap becomes considerable. Although the feeling of going TLE + WA on stage was rough, if I had prepared for so long and then missed the chance to present, I would definitely have regretted it.