Invited speakers
Šárka Zikánová (Charles University, Czech Republic): Text Structure and Its Ambiguities: Corpus Annotation as a Helpful Guide
The availability of digital language resources marks an important step forward in linguistic research, both for its theoretical as well as applicational orientation. The originally collected data gave an impulse to enrich them by various more sophisticated annotation systems dealing with most different phenomena and adding more levels of granularity.
Human data annotation is a process based on interpretation of observed phenomena. Human annotators may disagree in the evaluation of language expressions and structure; this variation may be seen as a negative feature lowering the quality of the data, which can be solved by unification of the output. In our talk, we follow a different approach, understanding the variation in annotation as an expression of a possible actual vagueness and ambiguity of language. We concentrate on the disagreement on understanding textual phenomena, such as discourse relations, coreference and information structure. The results of the analysis, i.e. identification of typical ambiguous points of the language structure, can serve as a basis for psycholinguistic experiments and e.g. for a later formulation of recommendations how to decrease ambiguity and increase intelligibility of administrative texts, schoolbooks, law texts etc.
Šárka Zikánová is a Czech linguist, a research associate at the Institute of Formal and Applied Linguistics (Charles University, Faculty of Mathematics and Physics, Prague, Czech Republic). She stayed on foreign scientific internships at the universities of Leipzig, Krakow, Philadelphia and Edinburgh. Her research interests evolved from the word order in Older Czech (position of the clitic se, position of a verb – a monograph in 2009, information structure, Latin influences) to the information structure and discourse relations in new Czech, compared to other languages. She is especially interested in the interplay of syntax, information structure, discourse relations and coreference (cf. a collective monograph of the Prague discourse team in 2015). She took part in the development of the Prague Dependency Treebank, the Prague Discourse Treebank, the Czech RST Discourse Treebank and other corpora. Her specific research area is implicit discourse relations (a monograph in 2021). Nowadays, she is turning her attention to the psycholinguistic studies of discourse and other coherence phenomena.
Ostap Okhrin (Technical University of Dresden): Reinforcement Learning in Transportation
Reinforcement learning (RL) has emerged as a powerful method for solving complex control tasks across various domains, from autonomous driving to maritime navigation. Work of my team in RL, particularly in value-based algorithms, addresses critical issues such as overestimation bias, proposing innovative solutions like the T-Estimator (TE) and K-Estimator (KE) for bias control and algorithmic robustness. Our advancements are validated through modifications to Q-Learning and the Bootstrapped Deep Q-Network (BDQN), demonstrating superior performance and convergence. Additionally, we have developed a spatial-temporal recurrent neural network architecture for autonomous ships, enhancing robustness in partial observability and compliance with maritime traffic rules. Our recent endeavors also include a modular framework for autonomous surface vehicles on inland waterways, utilizing DRL agents for path planning and following, significantly outperforming traditional control methods. Moreover, our work on dynamic obstacle avoidance environments for mobile robots and drones emphasizes the importance of controlled training difficulty for better generalization and robustness. This approach has been successfully applied across different platforms, reducing the simulation-to-reality (Sim2Real) gap and improving performance in real-world scenarios. Through these contributions, we aim to advance the practical application and reliability of reinforcement learning in diverse and dynamic environments.
Ostap Okhrin is Professor of Econometrics and Statistics, especially in Transportation, at the Institute of Transport and Economics, TU Dresden, Germany. He has co-authored nearly 100 publications in the field of mathematical and applied statistics, econometrics, and reinforcement learning, with applications to finance, economics and autonomous driving.
Primož Potočnik (University of Ljubljana, Slovenia): Generating and cataloging symmetric graphs
Symmetry is a concept which plays a significant role in
many areas of human activity. In mathematics, the desire to understand
symmetry gave birth to modern group theory. Even today, groups are
often studied in terms of their representations as symmetry groups of
fixed mathematical objects, such as graphs. The study of groups acting
on graphs, especially highly symmetric graphs, have resulted in many
deep theories and ground-breaking results throughout mathematics.
When studying a class of discrete objects, it is of profound
importance to be able to construct complete lists of the
representatives of this class up to a given size. On one hand, such
lists (or censuses, as are often called) help us formulate and test
conjectures. On the other hand, the lack of practical means of
constructing such censuses indicates the lack of understanding of the
theory. Thus, development of theory (and, of course, computational
resources) enables constructions of complete censuses of objects, while
censuses themselves facilitate and motivate further theoretical
achievements.
Attempts of constructing census of graphs with high level of
symmetry began in early 1930s, when Foster started collecting examples
of arc-transitive graphs of valence 3. His work, now known as Foster's
census, has been a valuable source of information for graph and group
theorists for many decades. Several legendary mathematicians have been
involved in constructions of catalogues of graphs of specific symmetry
types, such as William Tutte, Harold Coxeter, John Conway etc. In the
last few decades, Foster's original work was successfully upgraded in
several directions. The aim of the talk is to present the theory and
methods behind constructions of these classical censuses, give a few
practical demonstrations using modern computer algebra systems, and
present some of the more recent achievements in this area.
Primož Potočnik is a professor at the University of Ljubljana, the head of the Group for Computationally Intensive Mathematical Methods, and from 2022 to 2024 served as the president of the Society of Mathematicians, Physicists and Astronomers of Slovenia. He earned his Ph.D. from the University of Ljubljana in 2000 with a dissertation on Graph Symmetries and Group Actions. He was a postdoctoral fellow at the University of Ottawa, Canada, and a repeated visitor at the University of Auckland, New Zealand. His primary research focus is on the intersection of group theory and combinatorics. He is well-known for his involvement in the creation of several datasets of highly symmetric graphs and combinatorial objects. He collaborates frequently with other mathematicians, including a number of them from Slovakia, on various topics concerning arc-transitive and vertex-transitive graphs, and has been the doctoral adviser as well at the host for a number of postdoctoral visitors.
György Vaszil (University of Debrecen, Hungary): Describing computations of membrane systems with Petri nets
Membrane systems are biologically inspired models of computation: their operation attempts to imitate the functioning of living cells. The computation in a membrane system proceeds in distinct regions called membranes or compartments. The compartments allow computation with multisets: they accomplish transformations of their contained multisets by various evolution rules. In the original symbol object model, the compartments are organized in a tree like structure and the rules account for the computational processes in the compartments in a parallel and distributed manner. Multisets of objects on the left hand sides of rules are transformed to multisets on the right hand sides and these might also be placed to other regions of the system before the next computational step begins. The computation proceeds until no more rule can further be applied in any of the regions, and the result is usually given by the objects of one or more designated output regions.
When looking at the computations of membrane systems and the behavior of place/transition Petri nets, we might notice several features which are related to each other. Petri net transitions consume tokens from their input places and produce new tokens at their output places, so in some sense they behave similarly to membrane systems which consume, produce, and move objects around in the regions of their membrane structure. Based on these relationships, the functioning of place/transition nets can naturally be described by transformations of multisets corresponding to possible token distributions on the places of the net, while different kinds of objects and object evolution rules in different compartments of a membrane system can be represented by the places and transitions of a Petri net.
In the presentation we look in more detail at these structural links between the two models which, on one hand, motivate the examination of membrane systems from the point of view of the concurrent nature of their behavior, and on the other hand, inspires the study of Petri net variants suitable for the modeling of membrane system computations.
György Vaszil obtained his PhD in 2001 at the Eötvös Loránd University of Budapest. Since 2015, he is full professor at the Faculty of Informatics of the University of Debrecen, where he is the head of the Department of Computer Science. His research interests include the theory of formal languages and automata, unconventional or nature motivated computational models, such as bio-inspired models like membrane systems. He has published more than 150 papers in international journals, conferences, and workshops.
