As it turns out I handed in my thesis last Friday! I have quite mixed feelings about it. Part of me says that’s great, the other part says more like “but why didn’t you think of that?!”
Last few weeks were a bit crazy as I was working to get the measurements I wanted to support (or counter) some decisions. The report could certainly do with some more polishing, if not content, at least layout-wise. Making things pretty gives bonus points right?
Anyway, the last weeks’ hard work culminated in an internal presentation for the engineers at Tuenti. I tend to like doing presentations (and planning them). Although I ended up being very short of time for the preparation, I felt happy with the delivery. The slides are very sketchy (attached below) and I hope to refine them, and the content, for the final presentation this coming Friday.
Questions I remember:
- Why matrix factorisation?
- How could you address the router bottleneck?
- What is the memory and cpu utilisation?
- On what machines did you run the experiments?
- What clustering algoritm was used to cluster the items?
Comments received afterwards:
- More details about the system including the setup and the tools used (Scala+Akka)
- A bit too fast through the architecture
- Focus a little bit more on scalability
- There’s a simpler way to explain matrix factorisation (I’m all for making things simpler)
- (Own comment) Too slow through the motivation – cut down and focus on the most essential
This was the first presentation I did where part of the audience was available on video-link. It felt awkward and hard to connect with them. Also, I did the presentation sitting down (our video conference room is not great for standing remote presentations). Sitting, however, I will try to avoid in the future as that made me feel restricted.