It is not every day that a little, curiosity-driven side project of a bachelor student finds international scientific recognition. Yet, for second year Computer Science student Sebastian Bugge Loeschcke, this dream came true, and he will present his research results at the International EuroVis Workshop on Visual Analytics (EuroVA) at the end of May.
In the paper “Progressive Parameter Space Visualization for Task-Driven SAX Configuration”, Sebastian investigates one of the main questions in our data-driven world: How to turn big data into more manageable, smaller datasets without losing too much information? The answer to this question highly depends on what the dataset is used for. For example, a trend analysis may still produce reliable results when working on only every fifth data point, while some machine learning approaches may require every second data point to yield good enough results.
With Sebastian’s visualization software, the type of analysis as well as the desired accuracy of its results can be factored into the decision of how much data to remove. Once removed, the datasets are considerably smaller, which in some cases enables the analysis in the first place, or can speed up long-running computations in other cases. Sebastian has applied his method to reduce the size of so-called light curves, which are astronomical datasets used to identify exoplanets. He showed that reducing the data from the astronomical observations to a third of their original size still allows running the usual detection algorithm for exoplanets with reliable accuracy.
Getting to these results took tenacity and many hours of coding on top of Sebastian’s regular study work. He worked on this project for two semesters as a talent track activity. The talent track offers bachelor students the opportunity to come in close contact with current research topics being pursued at the department. It is a great way of learning the ropes of research in different areas of computer science at the beginning of your university studies. “From weekly meetings with my advisor and many chats with the PhD students in the hallway, the new data reduction approach emerged” says Sebastian. “I had to overcome many roadblocks, before I finally got the first useful results. But I had a lot of fun solving those problems that I could tackle, to think around the ones that were too hard to solve, and to pick-up some experience with useful visualization tools along the way,” he continues.
He shares this experience with a number of other students, who turned their master thesis research into scientific publications. Michael Reidun Engelbrecht Larsen, for example, got the outcome of his master thesis accepted as a short paper at this year’s EuroVis – Europe’s leading scientific visualization conference. He investigated the use of sketchy rendering to create more memorable information visualizations. Another example is a paper on a new layout algorithm for stacked area charts by Steffen Strunge Mathiesen, which recently was submitted to IEEE VIS, the highest-ranking international visualization conference. Steffen has been working on improving stacked area charts ever since last fall’s Data Visualization course. Now, he has been using his master thesis to further fine-tune and improve his ideas.
>>> SEE ALSO: Very Successful First Run of New Data Visualization Course
“These projects illustrate what our research-based teaching approach the department can achieve” says Assoc. Prof. Hans-Jörg Schulz, who advises the visualization students, and is responsible for the department’s talent track education. “All of our students learn the basics of scientific inquiry, of challenging the state-of-the art in computer science. And if some students’ curiosity yields notable, novel results, we’ll do everything to make sure they get the recognition they deserve for it.” For the students, a first peer-reviewed publication or presentation at an international scientific conference often paves the way to a successful application for a PhD program, or as Sebastian puts it: “Once you’ve experienced what it is like to do research in computer science, you definitely want to do more of it.”
