The First 'Visible' Breakthrough

It’s been a while since the last #phdlife blog post. During spring and summer I was too busy to write one, and openly speaking, somewhat ashamed of not having made it to ICML’24 either. Of course, that’s nothing to be ashamed of if your work is not a pile of nonsense, given the skyrocketing number of submissions and the existing deficits in the review system. Yet, it took me a successful publication to muster up the courage to post on this blog again. Having made peace with myself in the beginning of the year, me and my coauthors have adjusted our approach, and we have worked hard during spring and summer to prepare the next version of the paper. This time, we went for “Transactions on Machine Learning Research” (TMLR) because this one would place our fairly uncommon RL problem into a reviewing system that takes away the numbers pressure without compromising on the review quality. Our paper improved further as we worked towards TMLR, and this time our interactions with the reviewers were much more pleasant and fruitful than the last times.

Contrary to my expectation, I didn’t break out into euphoria when receiving the final acceptance notification for this paper, because this time the initial reviews where already positively inclined. Nevertheless, when our paper got accepted in TMLR (link), it filled me with a deep sense of accomplishment and boosted my morale for several weeks. From that point on I finally believed that I can finish my PhD successfully and that it’s “just” a matter of consistent work and patience for the remaining time. Admittedly, this one paper is not enough for me to graduate, and I am not yet close to my next submission. But at least for now, I have the opportunity to focus on scientific discovery and enjoy my learning journey instead of thinking about how to generate measurable scientific “output” every other day. Of course, in an ideal world, that should be on the mind of a PhD student all the time. But honestly speaking, current-day ML research is very competitive and fast-paced in virtually all sub-domains you could be working on, and so it’s almost impossible to do research without anxiously looking at your paper output. Therefore my current situation is all the more pleasant.