Manuel Ullmann studied computer science at the University of Erlangen from 2005 to 2012 (degree: Dipl.-Inf.). He has been a research assistant at the Chair of Information Science at the University of Regensburg from 2013 to 2020.
As part of his research, the URWalking pedestrian navigation system was further developed. in the associated web app, destinations on campus, such as lecture halls or offices, can be selected. The system under development then calculates a route, taking into account preferences, such as avoiding outside areas when it rains. Using eye-catching objects, so-called landmarks, users are then guided to their destination. Insteda od instructions such as “turn left after 500 meters”, it means, for example, “walk past the landmark ‘Kugel'”.
Chair of Information Science
At the chair for information science
In my lectures “Analyzing and visualizing with Python” and “Acquiring, preprocessing and analyzing sensor data”, programming with Python, NumPy, Pandas ans scikit-learn is dealt with. Using the practical example, the students learn to prepare data, to explore it and to train models that assign the acquired characteristics to given classes.
In my event “Acquisition, preprocessing and analysing of sensor data” we use mathematical formalisms to be able to make rational decisions automatically. If you want to put it that way, we design intelligent agents, even if this term is of course worthy of discussion, since they have neither free will nor a strong sense of humor.
In fact, the previous knowledge among the students is very different. I therefore try to explain everything from scratch and I am happy if there are many questions. It seems important to me that you don’t let formulas kill you at the beginning, but try to understand the ideas behind the formalisms. In my events I try to convey why I am enthusiastic about machine learning and what exciting opportunities result from it.
I believe that the most important professional competence is being able to familiarize yourself with new topics independently. Even if nobody wants to work with neutral networks in ten years’ time, an understanding of basic concepts will certainly help to familiarize yourself with new technologies.
I am convinced that there are still many exciting developments in this area, although many may be disappointed if it turns out that robots will not be able to clean up the apartment in the next few years.
In our URWalking project, we are working on a navigation system for the university campus that proposes routes that correspond to the natural course of paths that users prefer. Now that all lectures take place digitally, one can of course ask whether such a system is still needed at all. the underlying optimization process can certainly also be applied to other problems, for example:
What is the fastest way to find the link to the online lecture?
The data-driven optimization of routes is, for example, a topic for which there are various methods, such as evolutionary strategies, which are certainly also interesting for teaching in the field of AI.
During my studies at the chair for AI, I had already heard lectures from Mr. Ludwig, so I quickly became enthusiastic about this topic. Of course, studying will be easier if you also take a look at the books that the lecturers use to prepare. I think it’s really great that there are now so many videos on the internet, for example from MIT lectures, which you can watch completely free of charge.
Learning new things is always difficult for me. Since topics related to AI are so exciting, I bite my way through!
I find it fascinating when you suddenly find correlations in a data set that would never have occurred without a systematic evaluation. Our wedding guests had to listen to an hour-long Powerpoint lecture on how the automated interests of my wife and I can be worked out automatically based on the telegram messages we have sent over the past five years. Isn’t that exciting?
In our Optapeb project, we work, among other things, with the Chair of Clinical Psychology and Psychotherapy. The project is about treating anxiety disorders by confronting patients with their fears in a virtual environment. So that this does not become too scary, we measure the user condition with various sensors. We classify the data obtained, for example, if the fear becomes too great, we can send a recommendation to the therapist to give a short praise. Such collaborations are always a great way to learn about new applications.
I hope that AI will continue to be promoted according to the current hype and the benefits will be recognized, even if the development sometimes does not go as quickly as you might hope. Even though it is often frustrating to tell Alex that you want the lights on, I think current developments are still leading to many drastic changes.
I think it’s great that this project exists and I hope that it will lead to more collaboration between different research groups. So far, however, I have had little to do with it myself.
I have already given everything I wanted to get rid of, even if it was not always asked 😉
Mr. Ullmann, thank you for taking the time to answer these questions. We wish you a nice day!