- Home
- Institute
- People
- Research
- Applications
- Seminars and Events
- Library
- Doctoral Studies
- Jobs

Institute of Computer Science

The Czech Academy of Sciences

The Czech Academy of Sciences

"Simplicity is a great virtue but it requires hard work to achieve it and education to appreciate it. And to make matters worse: complexity sells better." - E. W. Dijkstra

*TK05020142* 2023 - 2025

The content of the project is the development of a software solution capable of meeting the requirements for predicting a variable baseline according to ČEPS standards for a wide range of technologies on the consumption side and on the production side. The software solution - thanks to the innovative baseline prediction procedure - will significantly increase the possibility of involving these technologies in the provision of support services. In the solution, we will use innovative deep learning methods of neural networks with regularization (Deep Learning - DL & Deep Neural Networks - DNN). The newly created and implemented DNN models will subsequently be practically validated for baseline prediction for specific entities from among flexibility providers.

**GA23-07074S**
[Registered results] 2023 - 2025

Symmetry of the human brain (and the lack thereof) has been a matter of prominent debate since the report of the left-hemispheric dominance of language by Broca in 1865. Functions including memory, perception, learning, spatial cognition, attention, emotion processing and motor skills show degree of hemispheric specialization, and disrupted brain anatomy and more recently connectivity asymmetry has been associated with neuropsychiatric disorders such as schizophrenia. However the origin, character and consequences of brain asymmetry are far from understood. Recently, the use of large datasets and graph theory has been identified among key future directions in brain symmetry research. The outstanding challenges include relating structural and functional symmetries, bridging the gap between global symmetries (such as bilateral symmetry) and local symmetries, providing normative description of brain symmetries, and relation of inter-individual differences in brain connectivity (a)symmetry to cognitive and clinical characteristics such as in aberrant lateralization in schizophrenia.

**GA23-07931S**
[Registered results] 2023 - 2025

Dynamical systems are mathematical models of change in space over time. Such systems may sometimes be extended into the transfinite, and such extensions can be used to prove asymptotic properties about the original finitary system. On occasion, transfinite methods are the only methods known for proving these results, and sometimes may even be necessary, in a way that can be made precise using mathematical logic. The aim of this project is to analyze such transfinite methods in order to understand the logical and computational information provided by them and extend their usefulness. We will employ cutting-edge proof-theoretic and ordinal-theoretic techniques, including formalizations in the context of second order arithmetic, where the strength of many results in mathematical analysis has been systematically gauged through the technique of reverse mathematics. Such techniques must be intimately entwined with a deep understanding of the target fields, including analysis, combinatorics, and potential application areas in engineering.

*NSFC-23-08* 2023 - 2024

Discerning the cause from effect is the aim of many scientific disciplines. In this project we will further develop different methods for detection of causality from experimental time series and applied them to discover the causes of variability of Eurasian winter air temperature, with special attention to its extreme values. The causality methods will be based on information theory, dynamical system theory and compression complexity, combining methods from mathematics, statistical physics and computer science. Among the potential causes of Eurasian winter climate variability we will investigate the changing Arctic sea ice content as well as large-scale modes of circulation variability such as the North Atlantic Oscillation or Pacific Decadal Oscillation. The expected results will help to understand the causes of Eurasian winter air temperature variability and its extremes and serve as basis for their forecasting.

*DAAD-23-01* 2023 - 2024

Transcranial magnetic stimulation (TMS) is a tool that is used regularly in experimental and clinical research, as well as for therapeutic and diagnostic purposes. Yet, little is known about the detailed processes that lead to the nervous and muscular responses. This causes considerable uncertainty in the interpretation of results, and large interindividual variability in treatment effectiveness. The goal of this interdisciplinary project is to create a computational model of nerve and muscle activity in response to TMS, which synthesises current models for the conversion of electromagnetic fields to nervous activity (expertise of Leipzig), and the propagation of activity in the central and peripheral nervous system (expertise of Prague). This will be achieved by simulating the complete communication pathway from neuronal activity in cortex, to motor evoked potentials (MEP) in the muscular periphery. The resulting model will be calibrated with existing data (Leipzig), and be subjected to detailed parameter studies with the aim to improve the use of TMS as a diagnostic tool.

*LQ100302301* 2023 - 2027

*CK04000150* 2023 - 2025

**AA - Filosofie a náboženství**
[Registered results] 2022 - 2025

The project will lift an important and up to now largely overlooked idealization in logical models of group knowledge and action - the extensional view of groups where a group is reduced to the set of its members. Consequently, groups change identity when their membership changes, any uncertainty regarding who is in a given group is ruled out and the structure of groups is not reflected. However, firms, informal teams or even loose crowds typically remain even if they gain or lose members, and the identity of all members is rarely common knowledge in any group of moderate size. In order to lift this idealization the project will study an intensional view of groups, which loosens the relationship between group membership and group identity and takes into account its algebraic or relational structure, and apply this view to questions in multiagent epistemic logic (distributed and common knowledge, perfect recall) and coalition logic (groups’ agentive powers). It will allow for more natural applications of logical models to questions of collective memory, action and responsibility.

*22-06414L* 2022 - 2025

Mathematical induction is one of the essential concepts in the mathematician's toolbox. Though, its use makes formal proof analysis difficult. In essence, induction compresses an infinite argument into a finite statement. This process obfuscates information essential for computational proof transformation and automated reasoning. Herbrand’s theorem covers classical predicate logic where this information can be finitely represented and used to analyze proofs and to provide a formal foundation for automated theorem proving. While there are interpretations of Herbrand’s theorem extending its scope to formal number theory, these results are at the expense of analyticity, the most desirable property of Herbrand’s theorem. Given the rising importance of formal mathematics and inductive theorem proving to many areas of computer science, developing our understanding of the analyticity boundary is essential.

*22-01137S* 2022 - 2024

Classical logic models reasoning about Boolean combinations of atomic propositions. Modal logics extend it by adding propositional connectives (called `modalities') to allow reasoning about the modes of truth, such as `necessarily’, `is allowed', or `is known'. Conversely, substructural logics relax assumptions on logical atoms to allow reasoning about other interesting objects such as constructive proofs, resources, or the degrees of truth. There are deep mathematical theories available for both classes of logics, which both aid their applications in mathematics, computer science, economics, linguistics, etc., and are of independent mathematical interest. This is, however, not the case for their combination, which hinders their development and application potential. The goal of the project is to advance three underdeveloped areas of substructural modal logics by creating general theories of algebravalued frames and logics with layered syntax and establishing the foundations of quantified substructural modal logics.

*22-02067S* 2022 - 2024

Nowadays, modern AI technologies based on deep neural networks, whose computation is demanding on energy consumption, are implemented in devices with limited resources (e.g. battery powered cellphones). In error-tolerant applications (e.g. image classification), the use of approximate computing methods can save enormous amount of energy at the cost of only a small loss in accuracy. AppNeCo is a basic research project of approximate neurocomputing, whose ambition is an original synergy of approximation and complexity theory of neural networks and empirical experience with the top design of high-performance approximate implementations of hardware circuits. Its goal is to develop complexity-theoretic foundations of approximate computation by convolutional neural networks (CNN) of bounded energy complexity for application domains specified by input space distributions. This knowledge will be used in designing new strategies for approximating components and learning algorithms of low-energy high-precision CNNs. The new methods will be tested on image processing tasks.

*22-16111S* 2022 - 2024

Propositional Dynamic Logic, PDL, is a well-known tool used in the logical analysis of discourse about action. Being based on classical logic, it cannot provide adequate formalization of discourse involving graded, vague and imprecise concepts. This project will develop and study versions of PDL more suitable for this task, so-called graded dynamic logics. We will determine the basic properties of the most natural kinds of graded dynamic logic and we will develop versions of graded PDL aiming at formalizing various philosophically relevant kinds of discourse; in particular, we will develop graded dynamic logics suitable for formalizing reasoning about collective agency and deontic aspects of action in situations involving graded notions, and for analyzing reasoning about probabilistic aspects of action. The project will thus contribute to an elaboration of formal methods applicable in the theory of action and applied ethics.

2022 - 2025

During the last two decades, substantial progress in modelling urban microclimate processes has been associated with newly developed models. For the model validation, the precise and valid meteorological data represent the necessary information. This project represents a platform for open discussion about micro-scale measurement campaigns and its necessity for a correct interpretation of measured and modelled results.

*SEP-210669451* 2021 - 2024

Modal logics are a family of formal systems based on classical logic which aim at improving the expressive power of the classical calculus allowing to reason about “modes of truth”. The aim of the present proposal is to put forward a systematic study of substructural modal logics, understood as those modal logics in which the modal operators are based upon the general ground of substructural logics, weaker deductive systems than classical logic. Our aim is also to explore the applications of substructural modal logics outside the bounds of mathematical logic and, in particular, in the areas of knowledge representation; legal reasoning; data privacy and security; logical analysis of natural language. This is a 4-year project in the framework H2020-MSCA-RISE-2020: Research and Innovation Staff Exchange.

*NU21-08-00432* 2021 - 2024

Schizophrenia is a chronic, severe and profoundly disabling disorder. For every 100 individuals with schizophrenia, only 1 or 2 individuals per year meet the recovery criteria, and approximately 14% recover over 10 years, with poor functional outcome for 27% of patients. There is an urgent need to develop predictive models of outcome to be applied in the initial stages of illness and thus optimize and intensify intervention programs to avoid an aversive outcome. Functional outcomes are difficult to predict solely on the basis of the clinical features, but Magnetic Resonance Imaging (MRI), particularly multi-modal, holds promise for improved stratification of patients. The aim of this project is to develop tools to predict the functional outcome of schizophrenia from neuroimaging, clinical and cognitive measurement taken early after the disease onset. To overcome the limitations due to high dimensionality of MRI data, we shall apply a combination of robust machine-learning tools, data-driven feature selection as well as theory-based constraint to key brain networks characteristics.

*TL05000008* 2021 - 2023

The aim of the proposed multidisciplinary project is to create a set of procedures and tools for a complex analysis of educational tests and to promote the implementation of proposed procedures and tools in school admission tests, school leaving examinations, and other educational assessments. The project will provide instant analysis of educational tests including a comparison of parallel test variants as well as a detailed analysis of between-group differences, e.g. in terms of gender or school type. The proposed set will facilitate the analysis of data from administered tests and support the development of new tests used to evaluate educational outcomes or to select applicants. It will also support evidence-based decision-making at the school level and inform national policy-making.

*21-32608S* 2021 - 2024

Current psychological theory provides complex description of mental functions and processes. It is generally accepted that mental functions have brain as their substrate, and that mental processes and states are reflected in brain activity dynamics. A rapidly developing area of brain research is the study of spontaneous brain activity with functional magnetic resonance imaging, allowing simultaneous measurement of activity dynamics of a plethora of brain networks. It has been suggested that during resting state condition, the brain explores its dynamic repertoire of possible states. However, the structure and dynamics of such exploration remains elusive. We propose to use a combination of data analysis techniques, simultaneous EEG/fMRI measurement of both spontaneous and richly stimulated mental activity and comparison with the ever-growing body of functional neuroanatomical knowledge to characterize the state repertoire and transition dynamics of spontaneous mental activity as observable by neuroimaging methods.

*21-09458S* 2021 - 2024

Decision procedures for predicate logical theories play an increasingly important role in computer science, especially in combination with Boolean satisfiability solvers, that is, in SAT modulo theory (SMT) solvers. While there is a vast amount of current research on decision procedures for integers, real numbers, arrays, and many other theories, there is almost no results for the case of real functions, although such functions play a fundamental role in many areas of computer science and mathematics. We conjecture that the reason for this situation is the difficulty of the problem which we propose to overcome by designing so-called quasi-decision procedures for real functions. A quasi-decision procedure relaxes the decision problem in such a way that it is not required to terminate in borderline cases where the satisfiability of the input formula changes under small perturbations of this formula. In many applications, such borderline cases are actively avoided, and hence quasi-decision procedures can solve precisely those cases that are important in such applications.

*21-14727K* 2021 - 2024

The concept of synchronization of nonlinear dynamical systems will serve as basis for development of mathematical methods and computer algorithms for detection and characterization of interactions and dependence in multivariate nonlinear time series. Directional links and causal relations will be quantified using the tools of information theory. The developed methods will be tailored to specific properties of scalp electroencephalogram (EEG). Overall structure of EEG synchronization, reflecting the functional integration within and across different spatial and temporal scales, will be classified in machine learning algorithms as a candidate method for description of brain states and their changes due to mental disorders. In particular, synchronization and its changes in EEG of depressive patients will be tested as predictors of antidepressant therapeutic efficacy. The developed methods will be applicable not only in analysis of electrophysiological signals in neurology and psychiatry, but generally in analysis of complex multivariate and multiscale signals.

*TO01000219* 2021 - 2023

The main aims of the international project are to: considerably improve spatial resolution and quality of the urban atmospheric environment assessment on the basis of state-of-the-art modeling, observation and data analysis technologies; improve and validate advanced modelling tools with focus on modelling of the turbulent flow in complex urban environment; improve methods for combination of observational and model data; compare environmental effects of the selected impactful urban policy measures in Prague and Bergen; equip public authorities with a set of tools to support urban governance: focus on air quality and thermal comfort.

The project TURBAN benefits from Norway grants and Technology Agency of the Czech Republic

*21-21762X* 2021 - 2025

The theory of graph limits is one of the most important recently emerged tools of discrete mathematics. It has led to breakthrough solutions of many old problems in extremal graph theory, theory of random graphs and in particular in connecting discrete mathematics to fields such as probability, real anf functional analysis and group theory. In the project, we will study the foundations of the limit theory of graphs, graph norms, and connections to mathematical models of statistical physics.

*21-03658S* 2021 - 2023

Psychometrics, as a field concerning psychological, health-related, educational and other behavioral measurements, is the disciplinary home of number of statistical data science methods. This project studies theoretical and computational aspects of psychometrics with the aim to propose estimation and detection methods superior to traditional ones. Project focuses on two psychometric topics (estimation of reliability and detection of differential item functioning) and their extensions to more complex designs. Project also provides software implementations as well as simulated and real data examples demonstrating usefulness and superiority of the proposed methods. The theoretical and computational aspects and innovations improving psychometric algorithms covered in this project may have broader impact on statistical data science.

*21-17211S* 2021 - 2023

Development of methods for effective description of complex systems is a growing area of interdisciplinary research at the junction of cybernetics, informatics, mathematics and theoretical physics, with application to a range of scientific disciplines including neuroscience, sociology, economics, genetics, and ecology. One of the key problems is the robust characterization of the structure of interactions within a system based on the multivariate time series. A common method for system representation is using correlation matrix of the variables, treated as graph and studied using graph-theoretical metrics, in comparison with random graphs of matched size and density. Recent results show multiple issues of such approach: correlation matrices do not capture faithfully the interactions, neglect higher-order dependences, insufficiently capture the predictive power in multivariate nonlinear models and lead to a bias in key graph metrics. We shall move beyond the state-of-art and our recent results towards a more general and robust complex system characterization. The main aim of the project is to bring theoretical and methodological advances in complex network research by tackling four key challenges outlined in the abstract.

*L100302151* 2021 - 2023

*L100302001* 2021 - 2023

*CZ.02.2.69/0.0/0.0/18_053/0017594* 2020 - 2023

The Institute of Computer Science of the Czech Academy of Sciences created a post-doctoral position for an excellent junior researcher from abroad. A research stay of a junior ICS researcher and a PhD student at a research institution abroad will take place. Two short-term visits of administrative or technical employees of the ICS at a research institution abroad will take place.

This project is co-funded by the EU.

*SS02030031* 2020 - 2026

The main goal of the project is to develop methods of air quality control, methods of identification of air pollution sources and their share in air pollution concentrations with a focus on current main problems of air quality and difficult quantification of different types of pollution. Consequently, model tools need to be developed to identify dispersion of air pollution, both with regard to current concentrations but also with a view to future expansion. Part of the research is also the development of laboratory methods for air quality evaluation, both methods of manual, isotopic analysis of elements in aerosol particle samples and methods of elemental analysis of aerosol particles. With regard to the impact on the health of the population, the impact of ultrafine particles will be evaluated at 5 localities in the Czech Republic, also with regard to external influences such as meteorological conditions. The project also includes estimation of the fraction of fog and icing in the total atmospheric deposition and the outputs will be used for quantification of the ozone effect. An interesting result of the project will also be the maps of phytotoxic doses of ozone for various plants. The impact of transport is apparent across the whole project, both on the health of the population and on the pollutant and greenhouse gas emissions. An unforgettable task of this project is the development of methodologies and emission factors used in the preparation of emission balances in relation to the international requirements of the EU and the UN Conventions. Also, data standards for reporting obligations introduced by Act 25/2008 Coll. will be developed, which will be an essential element of the subsequently developed comprehensive information system on air quality.

*Akademická prémie* 2020 - 2025

*MSM100302001* 2020 - 2023

Human thermal environment is in researchers’ interest for several decades. At the same time, with increasing computation power is possible to solve more complex problems. One of them, urban climate, demands on resolution raises the question, whether turbulence should be treated as a (partly) resolved or as a (complete) sub-grid scale process. Atmospheric modelling of urban areas usually considers a limited number of relevant processes, alternatively it is performed with coarser resolution which cannot properly describe the conditions in street canyons. PALM-4U is the first open-source meteorological model based on large-eddy simulation (LES) principle with implementation of the most of urban canopy related processes. The recent version of PALM-4U include Urban Surface Model (USM) and Radiative Transfer Model (RTM), necessary modules for calculation of MRT for biometeorology module. Those features open new challenges in street-level scale research of biometeorology.

*20-27757Y* 2020 - 2023

The project focuses on problems in the overlap of discrete mathematics and probability theory. We consider basic discrete structures: graphs, digraphs, trees, uniform hypergraphs, which in applied areas are used as abstract models for networks, population dynamics, etc. We will study randomly generated discrete structures from a theoretical perspective. Focusing on random variables which count the number of certain substructures (e.g., copies of a given graph), we seek answers to the following questions: what are their asymptotic distributions, how likely are events that these random variables deviate significantly from their expected values. Among the objectives of the project is to study the similarity between the random regular graph and the binomial graph by resolving the Sandwiching Conjecture of Kim and Vu; estimating the upper tail probability for small subgraph counts in sparse random graphs; determining the limit distribution of functionals (e.g., number of large matchings, maximum independent set) in random graphs and Galton-Watson trees.

**LM2018113**
[Registered results] 2020 - 2022

Research infrastructure serves for Czech contribution to particle physics research on experiments at Fermilab consists of experiments on which Czech physicists collaborate in Fermilab and of infrastructures of the Czech collaborating institutions. We work on the Fermilab's experiments NOvA, D0 and we plan to join a new experiment in next years contribute to its design and construction. In the Czech Republic it is a RCCPP computing farm and physics laboratory in FZU, cluster for artificial intelligence and neural networks algorithms in ICS and numerical and statistical computing servers at CTU. The whole infrastructure serves for particle physics experiments and for researchers for many years. The RI as top world research environment serves also for education of undergraduate and postgraduate students. Further service is support and implementation of novel statistical and non-statistical methods for data analysis based on artificial intelligence and machine learning, using specialized computing infrastructures located at FNSPE, Czech Technical University in Prague and Institute of Computer Science, CAS. It includes statistical pre-processing of neutrino data samples such as dimensionality reduction, nonparametric probability distribution estimates, statistical homogeneity testing, or signal reconstruction by means of artificial intelligence methods.

*734922 (8. RP EU)* 2017 - 2022

Networks are present in our lives in numerous different environments: to name just a few, networks can model social relationships, they can model the Internet and links between web pages, they might model the spread of a virus infection between people, and they might represent computer processors/sensors that have to exchange information. This project aims to obtain new insights into the behaviour of networks, which are studied from a geometric and computational perspective. Thereto, the project brings together researchers from different areas such as computational geometry, discrete mathematics, graph drawing, and probability. Among of the topics of research are enumerative problems on geometric networks, crossing numbers, random networks, imprecise models of data, restricted orientation geometry. Combinatorial approaches are combined with algorithms. Algorithmic applications of networks are also studied in the context of unmanned aerial vehicles (UAVs) and in the context of musical information retrieval (MIR). The project contains the work packages: “Geometric networks”, "Stochastic Geometry and Networks", “Restricted orientation geometry”, “Graph-based algorithms for UAVs and for MIR”, and “Dissemination and gender equality promotion”. The project connects researchers from 14 universities located in Austria, Belgium, Canada, Chile, Czech Republic, Italy, Mexico, and Spain, who will collaborate and share their different expertise in order to obtain new knowledge on the combinatorics of networks and applications.