Unexpectedly, a brilliant researcher passed away. Marco studied computer science at the University of Amsterdam and graduated with honours (cum laude) in 1995. After his graduation he went to Lugano in Switzerland to perform his PhD research in the Istituto Dalle Molle di Studi sull’Intelligenza Artificiale (IDSIA) in the area of reinforcement learning, under supervision of the well-known prof. Jürgen Schmidhuber. After his PhD graduation in 1999, he did a short post-doc in the Intelligent Systems Group from the University of Amsterdam. He became assistant professor in the year 2000 at Utrecht University, and joined the AI institute ALICE, at the Rijksuniversiteit Groningen as tenure tracker in 2007. His main interest was reinforcement learning. This is the type of learning where, e.g., a lazy instructor tells the intelligent agent after an hour of driving that ‘it was pretty bad today’ (or ‘good’, as Marco preferred). Such a learning paradigm is much more complicated than loss and gradient-based learning (back propagation), which enjoys detailed error information over all output dimensions. In reinforcement learning, an intelligent agent needs to find out itself what was good or bad in previous actions, taken under previously-perceived states of the world. With Martijn van Otterlo, he edited and coauthored the book “Reinforcement Learning: State-of-the-Art” (2012, 19 chapters, 630 pages, Springer). In February 2019, Marco and I went to Prague. It was a multi-conference (ICAART, ICPRAM and ICORES). We had our own lectures, in our own topics. I had time to join his presentation at ICPRAM. After his presentation, a crowd of young PhDs clustered around the speaker. Naturally, because of the content of his presentation. But also because of his personality: Friendly, accessible, authentic and definitely not like your average prof. This pattern confirmed what we saw here in Groningen. Marco attracted large numbers of bachelor and master student. He meticulously corrected their texts, smoking a cigarette on one of the tables outside the Bernoulliborg. More often than not, he lifted the students above themselves, regularly resulting in a joint publication of their thesis work.
During the past months he was increasingly tormented by demons conjured up in his mind. Regardless of this sad affliction, he tried to be a good researcher and teacher until the end. He was an intense, trustworthy, righteous person with a fascination for our research field, internationally renowned, loved by colleagues and students. He will be sadly missed.
Haren, 23 September, 2021
Lambert Schomaker
Note: On Google Scholar, Marco’s statistics are: h-index 37, number of citations 6539, best-cited publication: 720 citations. It is a little-known fact among researchers in computer science and artificial intelligence that their beloved Google Scholar profile page will be completely removed after a university has disabled the email address. Therefore, I made a snapshot of his current profile state. Undoubtedly Marco’s work will continue to attract attention.