On March 24, 2026 at 3 p.m., a seminar will be held in the seminar room MLIT–134/3–310:
- Tatevik Bezhanyan
Application of machine learning and computer vision methods for experimental data analysis in radiobiological studies
Abstract: The report presents the results of the development of algorithms based on computer vision and machine learning methods for analyzing experimental data in radiobiological research, as well as creation of web services for the automation data analysis. To analyze video data obtained in behavioral experiments on small laboratory animals in the “Morris Water Maze” test system, a web service was developed based on an algorithm which utilized computer vision methods and the YOLOv11 deep learning model. Here, one of the main problems are addressed: tracking and obtaining numerical characteristics of the object movement. A study of platform search strategies by rodents was conducted using unsupervised learning methods to determine the possible number of classes. To automate the radiation-induced foci analysis, the web service “MOSTLIT” was developed based on a two-stage deep learning approach. MOSTLIT allows the user to observe the identified nuclei, automatically detect two marked foci and obtain the numerical characteristics. Data annotation, algorithm development, and web service deployment were carried out using the ML/DL/HPC ecosystem of the HybriLIT Heterogeneous Computing Platform. - Dmitrii Kokaev
Mathematical Modeling of Hybrid Josephson Structures Consisting of Superconductors and Magnets
Abstract: The phase dynamics and resonance properties of the φ0 Josephson junction implemented in a superconductor-ferromagnet-superconductor hybrid structure are investigated. Algorithms for calculating the current–voltage characteristic (CVC) and constructing the voltage dependence of the magnetization component amplitudes in the φ0 junction are developed. It is shown that the amplitudes of magnetization components are modulated along the CVC. An approximate analysis demonstrates that this modulation is due to the implementation of parametric resonance in the φ0 junction. Additionally, the results are obtained from studying a system of three nanomagnets coupled to a Josephson junction will be presented.
Information on the seminar and the link to connect are available at Indico.