Filip Bartoš

Abstract:

This thesis focuses on solving PDE-constraint optimization problems using neural networks, specifically using physics-informed neural networks. We will present the results of applying this method to a particular physical problem that can be described by heat equation in 2D.

Štěpán Bezděk

Abstract:

This poster deals with procedural map generation for strategy games. It presents a comparison of several approaches to the algorithmic generation of game worlds and focuses in detail on the implementation of the height map method. It demonstrates the use of random generation, coarse grid interpolation and gradient noise (Perlin noise) to generate game terrain and presents visual results of generated maps with different terrain types.

Kryštof Blažek, Ing. Miroslav Kolář, Ph.D.

Abstract:

This bachelor project numerically investigates the propagation speed and convergence properties of SIR models. We use finite‐difference discretization in space and a fourth‐order Runge–Kutta scheme in time to simulate traveling‐wave solutions in one‐ and two‐dimensional domains. We perform an experimental order‐of‐convergence study under varying interpolation schemes (linear, natural spline, and not‐a‐knot spline) and verify the theoretical minimal wave speed derived via a supersolution argument. Additionally, we implement and test a moving‐boundary SIR model with Stefan‐type front dynamics.

Martin Jex

Abstract:

This study focuses on multiphase compositional equilibrium calculations in Volume-temperature-moles in each component $(VTN)$ and Pressure-temperature-moles in each component $(PTN)$ formulations, which are fundamental to a variety of applications in chemical engineering and thermodynamics, such as separation processes, reservoir simulations or ${CO}_{2}$ sequestration. We introduce novel enhancements to widely used algorithms and demonstrate their improved efficiency and robustness on several mixtures.

Pavel Ježek

Abstract:

The thesis explores the application of Generative Adversarial Networks (GANs) for predicting the prices of selected financial assets, specifically Apple stock, ČEZ stock, commodity gold, and the cryptocurrency Bitcoin. GANs, as an advanced deep learning framework, offer the ability to generate realistic time series predictions, which is particularly valuable for financial forecasting. The work covers the theoretical foundations of deep learning and GANs, a detailed description of the data, model implementation, and analysis of the results. Additionally, model-based trading strategies were applied to evaluate the practical viability of the predictions, including calculations of potential profits.

Patrik Kříž

Abstract:

This presentation showcases my thesis, which focuses on a practical use of Machine Learning models, namely Mask R-CNN and YOLO, and the creation of a dataset suitable for this problem. The aim of the thesis is to create a suitable model capable of counting the amount of people currently climbing and tracking which trails are the most used. This information is then going to be used in stopping overcrowding on the wall and helping the administrator know which trails are the most prone to erosion.

Ondřej Marek

Abstract:

This contribution studies the multi-speed entropic lattice Boltzmann method (ELBM). A generalization of a single-speed LBM wall boundary condition to multi-speed models is proposed and compared. A numerical study of drag and lift coefficients for a cylinder and NACA profile is perfomed using a single-speed LBM model, single-speed ELBM model, multi-speed ELBM model and the finite volume method.

Matěj Michálek

Abstract: This work presents the development of an advanced C++ software tool with a graphical user interface (GUI) designed to facilitate efficient communication with FPGA devices used in the COMPASS/AMBER experiments at CERN. The primary function of the tool is to provide intuitive and flexible access to the internal registers of FPGA-based hardware, simplifying tasks such as configuration, monitoring, and debugging. By abstracting low-level operations and presenting them through an accessible interface, the application enhances user interaction with the hardware and supports more effective system control in high-energy physics experiments.

Jakub Michna, Ing. Pavel Strachota Ph.D.

Abstract: TBA

Michal Moc

Abstract:

TBA

Maneesh Narayanan, Michal Beneš

Abstract:

We investigate the area-preserving flow of a closed embedded curve under constrained motion with force terms. Specifically, we analyze the deformation of a circle under two force scenarios: a droplet under external force and a circular-shaped eukaryotic cell. Reformulating the motion law as a system of degenerate parabolic PDEs, we solve it numerically using the finite volume method. Our findings offer insights into constrained geometric flows with applications in physical and biological systems.

Jan Oršl

Abstract:

This thesis focuses on the integration of a checklist, developed as part of a bachelor's thesis, into the existing logbook system to enhance efficiency and accuracy in task tracking. It compares different methods of checklist integration and their benefits, including error reduction and improved clarity of recorded information. Emphasis will be placed on the technical aspects of implementation as well as the user-friendliness of the solution. The thesis also includes the design of innovative logbook modifications that will expand their functional capabilities.

Roman Pirogov

Abstract:

In my work, I focus on deriving the weak formulation of the FitzHugh–Nagumo system of reaction-diffusion equations in a
two-dimensional domain. The aim is to find a weak solution in Sobolev spaces and to approximate it numerically using the finite difference method. The model simulates the propagation of electric signals in cardiac tissue. The results show how the system’s dynamics depend on model parameters and the domain type.

Neda Bagheri Renani

Abstract:

This work provides a detailed comparison between two commonly applied algorithms for solving nonlinear optimization problems: the enhanced Interior-Point approach (IPM) and the enhanced Inexact-Newton-Smart test (INS) approach. The enhanced INS method incorporates regularization of the Hessian matrix, iterative line search strategies, and improved termination rules to boost both numerical stability and the precision of the solution. While the IPM delivers more accuracy in optimal value approximation, the modified INS technique achieves increased convergence rates, computing efficiency, and numerical stability. The results show that the enhanced IPM method offers better computational performance and faster convergence while also achieving greater accuracy in meeting optimality criteria. By clarifying the computational properties of various methods, this work advances large-scale optimization by assisting in the selection of suitable techniques according to the needs of individual problems.

Daria Soboleva

Abstract:

Filip Šebek

Abstract:

This talk presents my thesis, which focuses on the mathematical modeling of the propagation of biochemical agents in heterogeneous media. The model is based on the 3D convection-diffusion equation, which is numerically solved using the finite difference method implemented in a custom code. The aim of the thesis is to simulate biomedical scenarios such as corrosion processes of biodegradable implants or drug delivery to tumor tissues.

Josef Štemberk

Abstract:

The theoretical part of the work connects concepts from elasticity theory with tools from functional analysis and continuum mechanics. It explores selected problems related to the behavior of elastic end elasto-plastic materials. The practical part focuses on the design of a solver for approximating the solution of a 1D static equilibrium problem for thermo-elastic material, based on the Finite Element Method.

Dalibor Trampota

Abstract:

In this talk, I'll walk you through a development of a CPU-based ray tracer I wrote as part of my academic work. I'll cover the core rendering pipeline including intersection algorithms like Möller-Trumbore and acceleration structures such as octee.

Martin Tůma

Abstract: This contribution explores the use of virtual reality (VR) for visualizing fluid flow simulations, focusing on the implementation of the Lattice Boltzmann Method (LBM) in Unity. LBM is a numerical method that enables efficient simulation of fluid dynamics by solving a discretized version of the Boltzmann transport equation on a lattice grid, making it well-suited for modeling complex flow behavior. In the context of this work, different categories of virtual environments are briefly introduced, including Virtual Reality (VR) and Extended Reality (XR), with an emphasis on their potential for scientific visualization. The practical component demonstrates how Unity's XR toolkit can be used to create an environment for viewing and interacting with 2D LBM simulation results. The goal is to provide an accessible and engaging platform for exploring fluid behavior, which can serve as a foundation for further development in scientific visualization using VR.