My Pentalogue for PhD Students
My PhD Journey
A few days ago, I found my git repository that contains most of the code I wrote and used for experiments during my PhD. It has a non-intuitive name: gpu-poster. How can a poster be converted into a PhD thesis?
It all started 6 years ago, when I was in my second year of PhD, without a clear idea of what my thesis will look like. There was a call-for-posters for a workshop sponsored by Nvidia. I decided to quickly run some experiments on a GPU, using the code I was working on but which was targeting CPUs.
I am not going to go into tech details. I just want to mention that I saw some potential in this idea and the poster was accepted. Next, I developed the poster into a paper that was initially rejected. However, I wrote a second paper where I analyzed the performance of different systems with GPU and this one was accepted by a good conference.
Within one year, my small repository had the implementation of a method to run analytics faster using GPUs, had around 8 benchmarks to test this method and scripts to run it on 4 different types of systems.
In six years, the core of the repository is the same. I just added more systems and a modeling tool to estimate the performance of my approach on different configurations. In six years, this repository generated 4 papers and 1 poster.
My Pentalogue for PhD Students
Be well organized. Keep your code clean, commented, and write some documentation on how to compile and run it. Keep ALL the log files of your experiments.
Read (a lot of) papers. And try to summarize each relevant paper in half a page or one page.
Write papers. For some of us, it is not easy to write. But it’s a good idea to keep trying. Try to write and submit your papers. You will get valuable feedback.
Create posters. Especially in the first two years, try to implement your ideas, create posters, and participate in conferences.
Discuss. Discuss your ideas and research, in general, with your colleagues, supervisors, friends, and strangers at conferences.
