The Machine Learning for Preventing and Combating Pandemics (MLPCP) workshop (virtually, May 8th, 2021), part of the International Conference on Learning Representations (ICLR) venue, offers researchers worldwide a venue to present and discuss their work combating pandemics. The Algorithm Engineering group is proud to contribute one paper, originating from the project seminar Competitive Programming with Deep Learning, to this year's edition. In the paper, the participants Otto Kißig and Martin Taraz present a drug repurposing strategy which performs significantly better than the previous state of the art and is also significantly faster.
The Algorithm Engineering group is also proud to announce that five papers (one of which stems from the Project Seminar on Computability and Learning Theory) got accepted at the Computability in Europe (CiE) conference (July 5-9, 2021, virtually in Ghent, Belgium). For years now, this conference series brings together aspects of computability and foundations of computer science as well as the interplay of these theoretical areas with practical issues. In their paper "Learning Languages with Decidable Hypotheses", the participants of the project seminar study how the requirement and the timing of the represention of (formal) languages via characteristic functions (that is, both positive and negative information must be specified) affects a learners learning capabilities. In another paper, Karen Seidel and Timo Kötzing present the complete map of pairwise learning power comparisons when learning in the limit from positive information with finitely many memory changes. Moreover, in a second paper, with additional co-author Ardalan Khazraei, they provide further results towards a better understanding of (strongly) non-U-shaped iterative learning in the context of binary labeled input data. Lastly, Vanja Doskoč and Timo Kötzing provide a thorough study of monotone learners, respectively, a normal-form for semantically witness-based learners in their papers, making an important step towards obtaining a complete map comparing the pairwise learning power of semantic learners.