19 August 2022,
by Adrián Javaloy, University of Saarland, Germany
Hi, I am Adrián, a third-year PhD Student at the University of Saarland, Germany. While Machine Learning has become ubiquitous to our society, often times these systems struggle when presented with data coming from different sources (for example, images and their respective captions). My research therefore focuses on understanding this phenomenon, as well as developing methods that can take full advantage of the diverse information coming from different sources.
Thanks to SICSA’s Saltire Emerging Research Scheme, I went on a two-month research visit to the University of Edinburgh, where I closely worked with Antonio Vergari. Antonio is not only an old friend, but a world-renowned expert in tractable probabilistic models—that is, models that can compute complex probabilistic queries exactly, and in a reasonable amount of time (in stark contrast with current Deep Learning approaches). I had a blast working with Antonio. Early on, we discussed and discarded some ideas we talked about before my visit (that’s how science works!), and started exploring other research directions which we both got excited about, and which we are going to keep working on in the following months.
I feel the most invigorating aspects of my visit were two. First, I felt a step forward in my approach to research by working as a senior researcher who proactively seeks for problems to solve, and how to tackle them alongside other potential collaborators. Second, I was forced to express my ideas clearly: explaining your research topic in thirty seconds is not easy, especially if the person in front of you has never worked on your topic.
More importantly, I got out of my comfort zone. During my visit, I joined the Turing AI Reasoning Workshop, as well as the AI Reasoning reading group, both laying way off my usual research. I also had the opportunity to talk about my current interests at Amos Storkey’s group (followed by a pint with this lovely group). Definitely, my main take-away is the people I met during my visit. Not only I was surrounded by a group of supportive students, but I had the fortune to discuss—and hopefully collaborate in the near future—with Siddharth N. about multi-modal learning, or with Frank Mollica about the acquisition of kinship terms, two outstanding researchers in their respective fields.
Albeit short, this visit has granted me the opportunity to set the foundations for future collaborations, as well as to get to know an incredible environment to which I hope to come back in the future.