Publications

Proceedings paper

10.14311/APP.2026.57.0254

This paper presents a 3D point-cloud measurement method for acquisition of static physical dimensions of people relevant for pedestrian dynamics and fire safety applications, such as shoulder breadth, maximum depth, and body-projection area. The method was validated beyond laboratory conditions through an initial field experiment and proved sufficient reliability and applicability for estimating the body size of pedestrians. Analysis of the measured body dimensions indicates that representing a body-projection area of a person without any accessories by an ellipse with the corresponding width and depth is a sufficiently precise approximation. Furthermore, the data show only weak linear dependence between maximum width and depth, i.e both dimensions need to be measured. This study aims to contribute to a discussion on interdisciplinary cooperation and the use of detailed anthropometric data in pedestrian movement and evacuation research.

Proceedings paper

10.5220/0014428400004061

Recently, a considerable amount of malware research has focused on the use of powerful image-based machine learning techniques, which generally yield impressive results. However, before image-based techniques can be applied to malware, the samples must be converted to images, and there is no generally accepted approach for performing this conversion. The malware-to-image conversion strategies found in the literature often appear to be ad hoc, with little or no effort made to take into account properties of executable files. In this paper, we experiment with eight distinct malware-to-image conversion techniques, and for each, we test a variety of learning models. We find that several of these image conversion techniques perform similarly across a range of learning models, in spite of the image conversion processes being substantially different. These results suggest that the effectiveness of image-based malware classification techniques depends more on the inherent strengths of image analysis techniques, as opposed to the precise details of the image conversion strategy.

Proceedings paper

10.4230/LIPIcs.STACS.2026.50

Given a planar graph, a subset of its vertices called terminals, and k ∈ ℕ, the Face Cover Number problem asks whether the terminals lie on the boundaries of at most k faces of some embedding of the input graph. When a plane graph is given in the input, the problem is known to have a polynomial kernel [Valentin Garnero et al., 2017]. In this paper, we present the first polynomial kernel for Face Cover Number when the input is a planar graph (without a fixed embedding). Our approach overcomes the challenge of not having a predefined set of face boundaries by building a kernel bottom-up on an SPR-tree while preserving the essential properties of the face cover along the way.

Article

10.1007/s10766-026-00821-x

We present an optimised distributed‐​memory algorithm for the unpreconditioned conjugate gradient (CG) method. We experiment with using the CG algorithm with the compressed sparse row (CSR), compressed sparse row 5 (CSR5), compressed sparse row 2 (CSR2) and symmetric compressed sparse row (SCSR) sparse matrix formats. For the CG algorithm, we present a CSR5‐​based and an SCSR‐​based distributed‐​memory sparse matrix‐​vector multiplication (SpMV) algorithm. We developed five distributed‐​memory CG implementations using different combinations of the mentioned sparse matrix formats and SpMV algorithms. We benchmarked our implementations on a distributed‐​memory computer cluster with Intel x86‐64 CPUs. For the maximum benchmarked number of compute nodes, the median performance of our optimised distributed‐​memory CG algorithm with CSR‐​based SpMV was 58 % greater than of the state‐​of‐​the‐​art PETSc library, and the median performance of our optimised distributed‐​memory CG algorithm with SCSR‐​based SpMV was 105 % greater than of the PETSc library.

Article

10.1016/j.geb.2026.02.006

Adaptivity to changing environments and constraints is key to success in modern society. We address this principle by proposing “incrementalized versions” of Stable Marriage and Stable Roommates , asking what the computational cost is of adapting an existing stable matching after agents’ preferences have changed. We additionally require that the new stable matching should not deviate too much from the old one. After formalizing these incremental versions, we provide a comprehensive computational complexity landscape of Incremental Stable Marriage and Incremental Stable Roommates . To this end, we exploit the parameters “degree of change” in the input (difference between old and new preference profile) and in the output (difference between old and new stable matching). We obtain both hardness and tractability results. In particular, with ties in preferences, both problems remain computationally intractable even under minimal preference changes, whereas with strict preferences, both become (fixed-parameter) tractable, regardless of the extent of preference changes.

Article

10.1016/j.comnet.2026.112188

Machine learning techniques provide powerful tools for analysis of encrypted network traffic. We present a novel set of features and a distance measure suitable for a wide variety of distance-based machine learning techniques useful in classification, clustering, and novelty detection in encrypted traffic flow data. The proposed features are given by distances between probability distributions of such characteristics as packet sizes or inter-arrival times. The proposed distance measure incorporates those features in a fractional lp-metric with p close to 0.1. The effectiveness of the distance measure is evaluated using an extensive labeled dataset containing 154 web services. The dataset was captured on the ISP backbone, and we present it as supplementary material. Application of k-nearest neighbors (kNN) algorithm in combination with the proposed distance measure and feature selection gives the classification accuracy of 91.0%. Comparison with a deep learning model shows that the kNN is competitive with state-of-the-art models. On a different publicly available dataset with a low number of classes, our approach reaches accuracy 99.6%, outperforming models presented in the literature. The benefits of novel features are further demonstrated using the LightGBM model, i.e. without relying on distance-based techniques. Besides direct classification, we have used the local outlier factor and rank-based detection algorithms to detect novel traffic flows. We show that their performance improves when using the proposed distance measure. Finally, to speed up traffic classification we compared the performance of kNN and Approximate Nearest Neighbors (ANN) algorithm, and applied the Affinity propagation clustering algorithm to select a suitable subset of kNN/ANN training points.

Article

10.3390/s26041297

Highlights What are the main findings? Multi-source data fusion (GFS predictors + station observations + static physiography) consistently improves 24 h 2 m air temperature forecasts relative to raw GFS across all spatiotemporal splits. The best operational configuration is TabPFN-KNN, achieving MAE = 1.26 degrees C in the most demanding regime (time = validation, space = validation), i.e., approximate to 24% lower error than GFS (1.66 degrees C). What are the implications of the main findings? High-resolution, spatially continuous near-surface temperature fields can be generated from routinely available forecast inputs and regional station networks, supporting field-scale agricultural decisions. The hybrid design (station-level learning + physiography-conditioned KNN propagation) provides a deployable pathway for superresolution services integrated into sensor infrastructures (e.g., SensLog/ALIANCE).Highlights What are the main findings? Multi-source data fusion (GFS predictors + station observations + static physiography) consistently improves 24 h 2 m air temperature forecasts relative to raw GFS across all spatiotemporal splits. The best operational configuration is TabPFN-KNN, achieving MAE = 1.26 degrees C in the most demanding regime (time = validation, space = validation), i.e., approximate to 24% lower error than GFS (1.66 degrees C). What are the implications of the main findings? High-resolution, spatially continuous near-surface temperature fields can be generated from routinely available forecast inputs and regional station networks, supporting field-scale agricultural decisions. The hybrid design (station-level learning + physiography-conditioned KNN propagation) provides a deployable pathway for superresolution services integrated into sensor infrastructures (e.g., SensLog/ALIANCE).Abstract Accurate field-scale meteorological information is required for precision agriculture, but operational numerical weather prediction products remain spatially coarse and cannot resolve local microclimate variability. This study proposes a data fusion superresolution workflow that combines global GFS predictors (0.25 degrees), regional station observations from Southern Moravia (Czech Republic), and static physiographic descriptors (elevation and terrain gradients) to predict the 2 m air temperature 24 h ahead and to generate spatially continuous high-resolution temperature fields. Several model families (LightGBM, TabPFN, Transformer, and Bayesian neural fields) are evaluated under spatiotemporal splits designed to test generalization to unseen time periods and unseen stations; spatial mapping is implemented via a KNN interpolation layer in the physiographic feature space. All learned configurations reduce the mean absolute error relative to raw GFS across splits. In the most operationally relevant regime (unseen stations and unseen future period), TabPFN-KNN achieves the lowest MAE (1.26 degrees C), corresponding to an approximate to 24% reduction versus GFS (1.66 degrees C). The results support the feasibility of an operational, sensor-infrastructure-compatible pipeline for high-resolution temperature superresolution in agricultural landscapes.

Proceedings paper

10.65109/SZOC7665

We study the problem of candidate control in participatory budgeting elections. Our focus is on two prominent sequential welfare-based rules---GreedyAV and GreedyCost---which are widely used in practice. Candidate control asks whether we can strategically modify the set of available candidates so as to either ensure that a preferred candidate p is selected (constructive control) or prevent p from being selected (destructive control). Since all variants of candidate control under the two rules we consider are known to be NP-hard, we analyze the problems through the lens of parameterized complexity and approximability. Under the first lens, we provide a comprehensive classification with respect to natural parameters such as the number of voters, the number of controlled candidates, and the number of distinct costs, as well as their combinations. Within the second perspective, we establish a tight approximability bound.

Proceedings paper

10.5220/0014996200004094

The objective of this paper is fourfold. First, we use rewriting logic to endow BPMN models with an execution semantics consistent with the action-oriented notation in BPMN. This contrasts fundamentally with works using state-based models. Second, we demonstrate the effectiveness of our approach by applying it to an example related to distribution layer protocols in hierarchical enterprise networks. Third, we use Maude modules to structure, execute, test and verify the rewriting logic theories capturing the semantics of our BPMN models. Maude modules are provided with an operational semantics and denotational semantics. The operational semantics allows to simulate the behaviour of BPMN processes by concurrent rewriting (sequential rewriting being a particular case of concurrent rewriting in rewriting logic). Fourth, we use the obtained results to verify the BPMN models of the specified protocols. The detection of undesirable properties is illustrated by a typical error encountered in BPMN models.

Proceedings paper

10.5220/0015189900004094

We demonstrate that an association of OntoUML/OCL with DEMO OCD models offers an appropriate framework for modelling both the structural and dynamic behaviour of routing protocols from an ontological perspective. The practical utility of this approach is validated through its application to a partial specification of OSPF, a standard intra-autonomous system routing protocol. Subsequently, we present a mapping approach for the conceptualization and implementation of DEMO OCD models using Web Ontology Language. The implementation, serialized in Turtle, is deployed in an Apache Jena Fuseki SPARQL server, allowing the specified protocol to function as a knowledge-based system, querying and inferring new knowledge from existing data.

Proceedings paper

10.1007/978-3-032-20291-8_41

This paper presents a standalone GUI Windows application for the parametric evaluation and visual comparison of macro-scale image data, with a focus on gunshot residue (GSR) analysis. Developed in C# using the .NET/WinForms framework and AForge.NET library, the software enables interactive segmentation, radial metric extraction, and real-time image overlay or RGB fusion. It supports rapid inspection of infrared or RGB images, automatic measurement of particle area, radial distance, and angular dispersion, and one-click export to CSV for further statistical analysis. A validation study using 18 infrared thermograms acquired at three shooting distances (15, 25, 50 cm) demonstrates statistically significant radial shifts in GSR patterns. Although tailored for forensic analysis, the system’s modular design allows straightforward adaptation to biomedical imaging tasks such as burn mapping or lesion tracking.

Article

10.1016/j.artint.2026.104563

Inspired by the famous Target Set Selection problem, we propose a new discrete model to simultaneously spread two opinions within a social network and perform an initial study of its complexity. Here, we are given a social network, a seed-set of agents for each opinion, two thresholds for each agent, a budget, and a number of rounds. The first threshold represents the willingness of an agent to adopt an opinion if the agent has no opinion at all, while the second threshold states the willingness to acquire a second opinion if the agent already has one. The goal is to add at most budget-many agents to the initial seed-sets such that the process started with these extended seed-sets stabilizes within the given number of rounds, with each agent having either both opinions or none. That is, our goal is to ensure that the spread of opinions is balanced. We show that the problem is NP-hard, and thus we study the problem from the perspective of parameterized complexity. In particular, we show that the problem is FPT when parameterized by the number of rounds, the maximum threshold, and the treewidth combined. This algorithm also applies to the combined parameter, the treedepth and the maximum threshold. Finally, we show that the problem is FPT when parameterized by the vertex cover number, the 3-path vertex cover number, or the vertex integrity of the input network alone. To complement our tractability results, we show that the problem is W[1]-hard with respect to a) the sizes of the initial seed-sets and the feedback-vertex set number combined, even if all thresholds are bounded by a constant, and b) the budget, the 4-path vertex cover number, and the feedback-vertex set number combined, even if every activation process stabilizes in at most 4 rounds.

Article

10.1016/j.parco.2025.103166

This article presents a comparative analysis of two storage approaches: a SAN disk array, exemplified by an HPE 3PAR device, and a software-defined storage cluster constructed with the Ceph software. The objective of this comparison is to ascertain whether a software-defined storage cluster built with commodity servers can achieve comparable performance to a SAN disk array with a similar hardware configuration. The configuration used identical numbers of components of matching speeds, capacities, and hardware generation from the same manufacturer. By relaxing some requirements on the software-defined storage, we were able to benchmark all RAID levels with corresponding replication and erasure code settings. The results revealed that 3PAR performed 31 times better for 4 KiB data block writes than Ceph. On the contrary, the Ceph cluster surpassed 3PAR by a factor of 1.4 in 16 MiB large-block reads. The differences are explained in the text based on the theory of operation of the two types of storage. We propose criteria for choosing the correct type of technology for individual use cases.

Article

10.1016/j.cnsns.2025.109324

We implement the geometric method proposed in [14,17,21] to analytically predict the sequence of bifurcations leading to a change of stability and/or the appearance of new periodic orbits in the secular 3D planetary three body problem. Stemming from the analysis in [1], we examine various normal form models as regards the extent to which they lead to a phase space dynamics qualitatively similar as that in the complete system. For fixed total angular momentum, the phase space in Hopf variables is the 3D sphere, and the complete sequence of bifurcations of new periodic orbits can be recovered through formulas yielding the tangencies or degenerate intersections between the sphere and the surfaces of a constant second integral of motion of the normal form flow. In particular, we find the critical values of the second integral giving rise to pitchfork and saddle-node bifurcations of new periodic orbits in the system. This analysis renders possible to predict the most important structural changes in the phase space, as well as the emergence of new possible stable periodic planetary orbital configurations which can take place as the mutual inclination between the two planets is allowed to increase.

Proceedings paper

10.4230/LIPIcs.ESA.2026.92

Pattern-avoiding permutations are a central object of study in both combinatorics and theoretical computer science. In this paper we design a data structure that can store any size-n permutation τ that avoids an arbitrary (and unknown) fixed pattern π in the asymptotically optimal 𝒪(n lg{s_π}) bits, where s_π is the Stanley-Wilf limit of π. Our data structure supports τ(i) and τ^{-1}(i) queries in 𝒪(1) time, sidestepping the lower bounds that hold for general permutations. Comparable results were previously known only in more restricted cases, e.g., when τ is separable, which means avoiding the patterns 2413 and 3142. We also extend our data structure to support more complex geometric queries on pattern-avoiding permutations (or planar point sets) such as rectangle range counting in 𝒪(lg lg n) time. This result circumvents the lower bound of Ω(lg{n}/lg lg{n}) by Pătraşcu (STOC 2007) for the general case. For bounded treewidth permutation classes (which include the above-mentioned separable class), we further reduce the space overhead to a lower order additive term, making our data structure succinct. This extends and improves results of Chakraborty et al. (ISAAC 2024) obtained for separable permutations via different techniques. All our data structures can be constructed in linear time.

Article

10.1109/TNSM.2025.3636557

Accurate network traffic forecasting is crucial for internet service providers to optimize resources, improve user experience, and detect anomalies. Until recently, the lack of large-scale, real-world datasets limited the fair evaluation of forecasting methods. The newly released CESNET-TimeSeries24 dataset addresses this gap by providing multivariate traffic data from thousands of devices over 40 weeks at multiple aggregation granularities and hierarchy levels. In this study, we leverage the CESNET-TimeSeries24 dataset to conduct a systematic evaluation of state-of-the-art deep learning models and provide practical insights. Moreover, our analysis reveals trade-offs between prediction accuracy and computational efficiency across different levels of granularity. Beyond model comparison, we establish a transparent and reproducible benchmarking framework, releasing source code and experiments to encourage standardized evaluation and accelerate progress in network traffic forecasting research.

Article

10.1007/s00224-026-10270-1

We study a geometric facility location problem under imprecision. Given n unit segments on the real line, each with one of k colors, the goal is to place a point on each segment such that the resulting minimum color-spanning interval is as large as possible. A minimum color-spanning interval is an interval of minimum size that contains at least one point from a given segment of each color. We prove that if the input segments are pairwise disjoint, the problem can be solved in O(n) time, even for segments of arbitrary length. For overlapping segments, the problem becomes much more difficult. Nevertheless, we show that it can be solved in O(n* log^2 n) time when k=2, by exploiting several structural properties of candidate solutions, combined with a number of advanced algorithmic techniques. Interestingly, this shows a sharp contrast with the 2-dimensional version of the problem, recently shown to be NP-hard.

Article

10.26493/1855-3974.3445.d9e

We study reflection-symmetric realisations of symmetric graphs in the plane that allow a continuous symmetry and edge-length preserving deformation. To do so, we identify a necessary combinatorial condition on graphs with reflection-symmetric flexible realisations. This condition is based on a specific type of edge colouring, where edges are assigned one of three colours in a symmetric way. From some of these colourings we also construct concrete reflection-symmetric realisations with their corresponding symmetry preserving motion. We study also a specific class of reflection-symmetric realisations consisting of triangles and parallelograms.

Proceedings paper

We propose the study of the diffusion of two secrets in a heterogeneous environment, that is, there are two directed networks over the same set of agents. Our model combines two group identification processes, each with independent desiderata. Specifically, we have two types of targets: constructive, where we aim to expose a given group of agents to a secret, and destructive, where we aim to prevent a given group of agents from being exposed to a secret. To secure these targets, we can employ different control actions: either deleting an agent or introducing a previously latent agent. We provide a complexity classification for all combinations of the aforementioned targets and actions. Since classical analysis mostly yields negative results, we adopt a parameterized perspective, considering (combinations of) natural parameters, such as the number of controlled agents or the number of rounds of the activation process. Taking a parameterized perspective, we provide several FPT and XP algorithms along with their respective lower bounds. This leads to a complete understanding of the boundaries between tractable and intractable cases.

Proceedings paper

Deep learning approaches for histopathological image analysis require large, well-annotated datasets, which are difficult to obtain due to the need for domain expertise and privacy constraints. This work addresses dataset limitations through synthetic data augmentation using Denoising Diffusion Probabilistic Models (DDPMs). We propose a unified framework that explores two complementary strategies: fully synthetic image generation and partially synthetic generation via inpainting. Our approach systematically compares pixel-space and latent-space diffusion models on our in-house histopathology datasets consisting of post-transplant heart tissue biopsies from the Institute for Clinical and Experimental Medicine (IKEM). We evaluate the quality of synthetic data using standard metrics (KID, FID, LPIPS) as well as expert pathologist assessment. The synthetic samples are subsequently employed for augmentation in segmentation pipelines to assess improvements in detecting underrepresented structures, particularly blood vessels. Our experiments demonstrate that synthetic data augmentation enhances recall (a substantial gain in medical imaging where sensitivity to fine structures is critical) in vessel detection (from 0.34 to 0.42) while maintaining overall segmentation capability, though with some trade-offs in precision. These findings indicate that systematic diffusion-based augmentation represents a promising approach to addressing class imbalance in histopathological datasets, particularly for rare anatomical structures.