dlcp2025:program
Differences
This shows you the differences between two versions of the page.
Previous revision | |||
— | dlcp2025:program [27/06/2025 20:35] (current) – [Program] admin | ||
---|---|---|---|
Line 1: | Line 1: | ||
+ | {{ : | ||
+ | ====== Program ====== | ||
+ | // | ||
+ | |||
+ | **The list of accepted reports.** | ||
+ | |||
+ | <color / | ||
+ | |||
+ | //If someone did not find themselves in the list, please inform us by email [[dlcp@sinp.msu.ru]]// | ||
+ | |||
+ | ===== Section 1. Machine Learning in Fundamental Physics ===== | ||
+ | |||
+ | ==== 36. Реконструкция энергии космических лучей ультравысоких энергий, | ||
+ | |||
+ | |||
+ | //**М.Ю. Зотов** (1), А.А. Трусов (2) \\ (1) НИИЯФ МГУ, (2) Физический факультет МГУ// | ||
+ | |||
+ | Мы рассматриваем задачу реконструкции энергии космических лучей (КЛ) ультравысоких энергий по данным флуоресцентного телескопа EUSO-TA. Данные были собраны в 2015 г. на сайте эксперимента Telescope Array (ТА). EUSO-TA -- это небольшой телескоп-рефрактор с диаметром линз 1 м, полем зрения 10х10 градусов и временным разрешением 2.5 мкс, созданный для наземных тестов аппаратуры, | ||
+ | |||
+ | ==== 40. Фильтрация ложных максимумов ШАЛ с помощью нейросетевых методов в эксперименте СФЕРА-3 ==== | ||
+ | |||
+ | // | ||
+ | |||
+ | Проектируемый в настоящее время телескоп СФЕРА-3 предназначен для изучения космических лучей в диапазоне энергий 1–1000 ПэВ методом регистрации излучения Вавилова–Черенкова, | ||
+ | |||
+ | ==== 71. Использование нейронного автокодировщика для генерации показаний поверхностных детекторов Telescope Array ==== | ||
+ | |||
+ | // | ||
+ | |||
+ | Исследована возможность применения автоэнкодера (АЕ) для обнаружения аномалий в данных широких атмосферных линий (ШАЛ), смоделированных методом Монте-Карло. АЕ обучался исключительно на событиях с протоном в качестве первичной частицы, | ||
+ | |||
+ | ==== 52. Графовая нейронная сеть с механизмом внимания для кластеризации треков частиц по событиям в эксперименте SPD на ускорителе NICA ==== | ||
+ | |||
+ | // | ||
+ | |||
+ | Данная работа посвящена разработке методов глубокого обучения для кластеризации треков элементарных частиц по событиям. В данной работе рассматривается архитектура графовой нейронной сети с механизмом внимания (GANN) для классификации треков по событиям в каждом временном срезе на эксперименте SPD. В работе представлен новый подход к сортировке треков, | ||
+ | |||
+ | ==== 69. Machine Learning Approach for Lattice Quantum Field Theory Calculations ==== | ||
+ | |||
+ | // | ||
+ | |||
+ | In modern quantum field theory and statistical physics, the expectation values of observables are represented as integrals over function space. In most interesting problems, such integrals can only be computed numerically using lattice approximations, | ||
+ | |||
+ | ==== 75. Нейросетевое моделирование оптических солитонов, | ||
+ | |||
+ | // | ||
+ | |||
+ | В работе рассматривается моделирование распространения импульсов в оптической нелинейной среде с использованием обобщенного нелинейного уравнением Шредингера (ОНУШ) шестого порядка производной и с нелинейностью седьмого порядка. Проводится исследование несколько модификаций PINNs (гиперпараметры, | ||
+ | |||
+ | ==== 65. Temporal difference modulated spiking actor learning ==== | ||
+ | |||
+ | //**Yunes Tihomirov** (1), Roman Rybka (2), Alexey Serenko (2), Alexander Sboev (2) \\ (1) National Research University Higher School of Economics (HSE), (2) National Research Center Kurchatov Institute// | ||
+ | |||
+ | While neuromorphic computing offers substantial energy savings via spiking neural networks (SNNs), developing effective methods suited for hardware deployment for reinforcement learning in SNNs remains a challenge. We present novel spiking neural network architecture for the actor part of the actor-critic framework. The proposed approach incorporates a two-layer network trained using temporal difference modulated spike-timing dependent plasticity (TD-STDP). Evaluated on the classic Acrobot and CartPole control tasks, our SNN-based actor demonstrates competitive performance. Using local plasticity learning rules is important for future implementation on neuromorphic hardware. | ||
+ | |||
+ | ==== 41. SBI в задачах анализа динамических изображений многоканального детектора ==== | ||
+ | |||
+ | // | ||
+ | |||
+ | Традиционно байесовские модели, | ||
+ | |||
+ | ==== 76. Simulation of trawl processes using SINN architectures ==== | ||
+ | |||
+ | //**Belkova Kseniia**(1), | ||
+ | |||
+ | This work proposes an approach to simulating trawl processes using Statistics-Informed Neural Networks (SINN) — a stochastic counterpart to Physics-Informed Neural Networks (PINN). Trawl processes are a special case of ambit processes, which are used to model a broad class of spatio-temporal phenomena. These processes are defined via integrals over Lévy bases on moving sets, allowing for the modeling of various dependency structures in time series. A notable special case is the Gaussian Ornstein–Uhlenbeck process, which has an analytical representation. However, existing modeling methods are limited to a narrow class of trawl processes due to computational complexity, especially when the Lévy basis distribution does not admit a closed-form expression. The main result of this work is a training scheme for SINN based on the characteristic functions of the process’s finite-dimensional distributions. Unlike the original SINN training framework, the proposed approach does not require external simulation of the process during training. The effectiveness of the method is demonstrated on trawl processes, including the Ornstein–Uhlenbeck process, and compared with existing approaches. To validate the method, we used data from the Met Office MIDAS archive of land and marine weather stations: weather conditions for 2013 were modeled based on weather data from 2012. | ||
+ | |||
+ | ==== 39. ML-Based Optimum Sub-system Size Heuristic for the GPU Implementation of the Tridiagonal Partition Method ==== | ||
+ | |||
+ | //**Milena Veneva** (1) \\ (1) RIKEN Center for Computational Science, R-CCS, 7-1-26 Minatojima-minami-machi, | ||
+ | |||
+ | The parallel partition algorithm for solving systems of linear algebraic equations (SLAEs) suggested in [1] is an efficient numerical technique for solving SLAEs with tridiagonal coefficient matrices which consists of three stages. It works by splitting the original matrix into smaller sub-matrices and solving these smaller SLAEs in parallel. Originally designed for use with a large number of processors, this algorithm was implemented using MPI (Message Passing Interface) technology in [1]. The development of HPC applications typically consists of two key phases: writing code that functions correctly and then optimizing that code to enhance performance. The nature of the parallel partition method is such that the initial SLAE with N unknowns is partitioned into a number of sub-systems with m unknowns each. The size of the SLAE N that the user solves is usually determined by the size of the problem they need to solve, while the size of the sub-system within the parallel partition method m is a parameter that needs to be tuned. We present one of the optimizations made to our CUDA [2] implementation, | ||
+ | |||
+ | [1] Austin, T.~M., and Berndt, M., and Moulton, J.~D., A Memory Efficient Parallel Tridiagonal Solver, Preprint LA-VR-03-4149, | ||
+ | [2] NVIDIA, NVIDIA CUDA C++ Programming Guide. https:// | ||
+ | [3] Fix, H., and Joseph, L., Discriminatory Analysis. Nonparametric Discrimination: | ||
+ | |||
+ | ==== 44. Natural Image Classification via Quasi-Cyclic Graph Ensembles and Random-Bond Ising Models with Enhanced Nishimori Temperature Estimation ==== | ||
+ | |||
+ | // | ||
+ | |||
+ | Recent advances have demonstrated the effectiveness of spectral clustering on the beta-Hessian of Graham matrices constructed from quasi-cyclic graphs in the context of Random-Bond Ising Models (RBIMs). Notably, at paper [1] showed that combining LDPC-inspired graph with VGG16-extracted features from GAN-generated two-class images (e.g., dog vs. cat) significantly outperformed Erdős–Rényi baselines in clustering accuracy, improving overlap from 73.21% to 90.60%—and up to 93.23% when using cosine similarity [2]. In this paper, we extend these insights to natural multi-class datasets, specifically ImageNet-10 and ImageNet-100. We introduce a refined approach to estimating the Nishimori temperature and propose a mixture-of-graphs model built from an ensemble of optimized RBIMs. These models leverage diverse quasi-cyclic graph families—including Spherical graphs and Multi-Edge Type LDPC graphs—to create sparse, expressive interaction structures. Feature embeddings are extracted from a lightweight MobileNetV2-based CNN, compressing 1280-dimensional activations to 32–64 feature maps per image. Using ensembles of 3 to 9 graph models, our approach achieves classification accuracies of up to 98.7% on ImageNet-10 and 82.5% on ImageNet-100 under optimal conditions with 32-dimensional embeddings. We demonstrate that significant parameter reduction in the MLP classification head (from 1280 to 32) improves both computational efficiency and robustness to feature puncturing. Furthermore, | ||
+ | |||
+ | [1] Usatyuk, V.S., Sapozhnikov, | ||
+ | [2] Dall' | ||
+ | |||
+ | ==== 43. Analysis of the TAIGA-HiSCORE Data using the Latent Space of Autoencoders ==== | ||
+ | |||
+ | //**Yu. Dubenskaya**(1), | ||
+ | (1) SINP MSU, (2) IIAP NAS RA, (3) IPA IGU // | ||
+ | |||
+ | The aim of extensive air shower (EAS) analysis is to reconstruct the physical parameters of the primary particle that initiated the shower. The TAIGA experiment is a hybrid detector system that combines several telescopes and arrays of detector stations to record and analyze EAS data. At present, data from the telescopes and the detector station arrays is analyzed by deriving different sets of auxiliary parameters related to the physical features of the recording hardware. These sets of parameters are chosen empirically, | ||
+ | |||
+ | This study was supported by the Russian Science Foundation, grant no. 24-11-00136. | ||
+ | |||
+ | ==== 81. Проблема аугментация данных атмосферных черенковских телескопов в стерео режиме на примере установки TAIGA-IACT ==== | ||
+ | |||
+ | // | ||
+ | (1) SINP MSU, (2) IIAP NAS RA, (3) IPA IGU // | ||
+ | |||
+ | Изучение источников гамма-излучения высоких энергий (более 1 ТэВ) во Вселенной возможно только с использованием наземных установок большой площади для регистрации широких атмосферных ливней (ШАЛ). Регистрация ШАЛ осуществляется детекторами заряженных частиц и/или детекторами черенковского света. По данным этих детекторов необходимо определить направление прихода, | ||
+ | Современные тенденции в обработке больших данных в области гамма-астрономии методами машинного обучения показывают, | ||
+ | Аугментация данных путем простого вращения изображений атмосферных черенковских телескопов (АЧТ) продемонстрировала свою эффективность для задач в монорежиме. В ряде случаев такой метод может быть интерпретирован как наблюдение ШАЛ с другого положения в пространстве. Однако, | ||
+ | В данной работе рассматривается возможность аугментации данных АЧТ путем вращения положений телескопов вокруг оси ШАЛ для обучения нейросетевых моделей при наблюдениях в стереорежиме. | ||
+ | |||
+ | Работа выполнена при финансовой поддержке Российского научного Фонда, грант 24-11-00136. | ||
+ | |||
+ | ==== 42. Возможность применения метода нормализующих потоков для извлечения редких гамма событий в эксперименте TAIGA ==== | ||
+ | |||
+ | // | ||
+ | (1) SINP MSU, (2) IIAP NAS RA, (3) IPA IGU // | ||
+ | |||
+ | Среди многих методов исследования процессов, | ||
+ | |||
+ | Работа выполнена при финансовой поддержке Российского научного фонда, грант № 24-11-00136 | ||
+ | |||
+ | ==== 95. Гамма-астрономия ультравысоких энергий и проект TAIGA-100 ==== | ||
+ | |||
+ | // | ||
+ | |||
+ | ---- | ||
+ | 14 | ||
+ | |||
+ | ===== Section 2. Machine Learning for Environmental Sciences ===== | ||
+ | |||
+ | ==== 15. Восстановление приповерхностной влажности атмосферы над океаном с применением методов машинного обучения ==== | ||
+ | |||
+ | //**С. А. Вострикова** (1), М. А. Криницкий (1,2), С. К. Гулёв (2), М. П. Александрова (2) \\ (1) Московский физико-технический институт, | ||
+ | |||
+ | Влажность воздуха в приповерхностном слое атмосферы над океаном является ключевым климатическим параметром, | ||
+ | |||
+ | ==== 16. Сравнение моделей машинного обучения в задаче идентификации аномалий в данных визуальной съемки поверхности моря ==== | ||
+ | |||
+ | // | ||
+ | |||
+ | Обнаружение морского макромусора является важной задачей для защиты экосистем океана, | ||
+ | |||
+ | ==== 45. Application of Convolutional Neural Networks for Upper Ionosphere Remote Sensing Using All-Sky Camera Data ==== | ||
+ | |||
+ | //**Andrei Vorobev** (1), Gulnara Vorobeva (2) \\ (1) The Geophysical Center of the Russian Academy of Sciences, (2) Ufa University of Science and Technology)// | ||
+ | |||
+ | This study proposes an original approach to the automatic classification of the upper ionosphere state through machine identification of images captured by sky cameras, also known as all-sky imagers. Based on 10 years of sky observations within the auroral oval (Kola Peninsula, Russia), represented by 163,899 images with a 10-minute sampling interval, an intelligent information system was developed using convolutional neural networks. This system identifies whether an input image belongs to one of seven predefined classes and subsequently interprets the result. The analysis of performance metrics for the system, built on the ResNet50 neural network architecture, | ||
+ | |||
+ | ==== 54. Foundation models of ocean and atmosphere in 2025: milestones and perspectives. ==== | ||
+ | |||
+ | // | ||
+ | |||
+ | Over the past two years, large-scale deep-learning foundation models have evolved from atmospheric-only emulators into first-generation, | ||
+ | |||
+ | ==== 60. Сравнение моделей машинного обучения в задаче идентификации аномалий в данных визуальной съемки поверхности моря ==== | ||
+ | |||
+ | // | ||
+ | |||
+ | Обнаружение морского макромусора является важной задачей для защиты экосистем океана, | ||
+ | |||
+ | |||
+ | ==== 56. Detection of Irminger Rings in high resolution ocean hydrodynamic modeling data using artificial neural networks ==== | ||
+ | |||
+ | // | ||
+ | (1) Shirshov Institute of Oceanology, Russian Academy of Sciences, (2) Moscow Institute of Physics and Technology // | ||
+ | |||
+ | Deep convection in the Labrador Sea is a key component in the formation of the lower branch of the Atlantic Meridional Overturning Circulation (AMOC). It is known that mesoscale eddy activity in the Labrador Sea, represented by Irminger Rings (IR), influences the convection process. In order to analyze the impact of IRs on the spatial-temporal variability of the mixed layer depth, it is necessary to create a trajectory database of eddy motion, which poses the problem of IRs detection and tracking with high accuracy. In this study, we propose the novel technique for detection of IRs in high-resolution ocean numerical simulation. The research is based on the regional model of the Subpolar North Atlantic NNATL12. There are known automated eddy identification methods that are widely used as a tool for studying eddy activity in statistically significant samples. The most commonly used local extrema search method depends strongly on a number of parameters chosen by an expert exploiting this approach. In order to alleviate the subjectivity issue, we first implemented the automatic identification scheme for IRs based on the local extrema search. We optimized the scheme employing Bayesian optimization framework resulting in optimal values of the hyperparameters of this eddy identification algorithm. While the optimization significantly improved the quality of the identification, | ||
+ | |||
+ | ==== 73. Моделирование турбулентного переноса примесей в планетарном пограничном слое с применением методов крупных вихрей и методов машинного обучения ==== | ||
+ | |||
+ | //**И. А. Герасимов** (1), М. А. Криницкий (1, 2), Е. В. Мортиков (3,4) \\ | ||
+ | (1) Московский физико-технический институт (национальный исследовательский университет), | ||
+ | |||
+ | Планетарный пограничный слой атмосферы (ППС) играет ключевую роль в контексте переноса загрязняющих веществ, | ||
+ | [1] Hendrik Tennekes; The atmospheric boundary layer. Physics Today 1 January 1974; 27 (1): 52-63. [[https:// | ||
+ | |||
+ | ==== 51. Нейросетевое пространственное масштабирование полей приповерхностного ветра над Баренцевым и Карским морями ==== | ||
+ | |||
+ | // | ||
+ | |||
+ | В настоящем исследовании изучается возможность применения искусственных нейронных сетей к задаче масштабирования приповерхностного ветра над Баренцевым и Карским морями. Используются различные конфигурации модели глубокого обучения с пропускными соединениями, | ||
+ | |||
+ | ==== 82. Сравнение методов машинного обучения для учета связей с запаздыванием при моделировании городского острова тепла. ==== | ||
+ | |||
+ | // | ||
+ | |||
+ | В работе исследуются методы машинного обучения для аппроксимации разницы температур между городской и сельской местностью (интенсивности городского острова тепла) на примере Москвы и Санкт-Петербурга. В роли предикторов выступают долгосрочные осредненные по региону данные наблюдений с загородных станций и данные глобального реанализа ERA5 с шагом сетки 0.25° с 2012 по 2023 гг. Особенностью метеорологических данных является наличие сильной автокорреляции (связей с запаздыванием). Для учета этих зависимостей исследуются два подхода: | ||
+ | ==== 72. Deep Learning-Based Estimation of wind induced waves parameters from X-Band Radar Imagery ==== | ||
+ | |||
+ | // | ||
+ | |||
+ | Shipborne navigation radars are essential for safe maritime navigation, detecting vessels and obstacles. Reflections from the sea surface—known as Bragg scattering—produce sea clutter, typically filtered out as noise. When the sea surface is rough enough (wind speed > 3 m/s, significant wave height (swh) > 0.5 m), this clutter becomes visible in unfiltered radar images, allowing for the retrieval of wind-induced ocean wave parameters. Traditional wave parameter estimation relies on three-dimensional Fourier analysis and linear dispersion relationships, | ||
+ | |||
+ | |||
+ | ==== 83. USING MACHINE LEARNING METHODS FOR JOINT PROCESSING OF DATA FROM MULTIPLE SEMICONDUCTOR GAS SENSORS ==== | ||
+ | |||
+ | //**Isaev I.V.** (1,2,3), Chernov K.N. (4), Dolenko S.A. (1), Krivetskiy V.V. (2, 5) \\ (1) D.V. Skobeltsyn Institute of Nuclear Physics, M.V. Lomonosov Moscow State University, (2) Scientific-Manufacturing Complex Technological Centre, (3) MIREA – Russian Technological University, (4) Physics Department, M.V. Lomonosov Moscow State University (5) Chemistry Department, M.V. Lomonosov Moscow State University// | ||
+ | |||
+ | This study addresses the problem of environmental monitoring of air in cities and industrial areas, which consists in detecting gases and volatile organic compounds using semiconductor gas sensors. To provide selectivity in the detection of certain gases, as well as high temporal resolution of the sensors, nonlinear temperature operating conditions were used - the so-called heating dynamics. Due to high complexity of physical and chemical models describing the processes of interaction between gases and sensors, machine learning methods based on the use of physical experiment data were used to process the sensor response. To provide additional selectivity in the detection of specific gases, this study considers simultaneous use of data from multiple semiconductor sensors with various doping components with building machine learning models capable of providing joint processing. Based on the results of the study, conclusions were made regarding the selection of optimal combinations of sensors and heating dynamics for a specific gas/all gases. \\ | ||
+ | The study was carried out at the expense of the grant No. 22-19-00703-P from the Russian Science Foundation. | ||
+ | |||
+ | ==== 67. Доменная адаптация нейронных сетей в задаче диагностики природных вод по спектрам комбинационного рассеяния света ==== | ||
+ | |||
+ | // | ||
+ | |||
+ | Современная экологическая ситуация характеризуется возрастающим антропогенным воздействием на природные жидкие среды, в частности, | ||
+ | |||
+ | ==== 89. Восстановление высоты зданий с использованием машинного обучения и цифровой модели поверхности ArcticDEM ==== | ||
+ | |||
+ | // | ||
+ | |||
+ | Моделирование погоды и климата для урбанизированных территорий требует детального описания городской подстилающей поверхности. Стремительное преобразование городской среды требует регулярного обновления необходимых для расчета данных. Несмотря на появление все большего количества моделей городского климата в разных масштабах, | ||
+ | |||
+ | ==== 92. Intercomparison of machine learning approaches for identifying hail from basic weather parameters ==== | ||
+ | |||
+ | //**Blinov P.D.** (1), Chernokulsky A.V. (2), Krinitsky M.A. (3, 4), Bugrimov S.A. (5) \\ (1) National Research University - Higher School of Economics, Moscow, Russia, (2) A.M. Obukhov Institute of Atmospheric Physics RAS, Moscow, Russia, (3) P.P. Shirshov Institute of Oceanology RAS, Moscow, Russia (4) Moscow Institute of Physics and Technology, Russia (5) Lomonosov Moscow State University, Moscow, Russia// | ||
+ | |||
+ | This work presents an integrated approach to hail diagnosis using ERA5 reanalysis data and Russian ground observations. We investigate the efficacy of three distinct methodologies: | ||
+ | |||
+ | ==== 55. Enhancing the Quality of Kp Index Machine Learning Forecasting Using Higher-Frequency Data and Feature Transformations | ||
+ | |||
+ | // | ||
+ | |||
+ | In this study, we investigate the problem of increasing the quality of existing models for Kp index forecasting up to 24 hours ahead with hourly step. We show that one way to do so is to incorporate 5-minute frequency data on the parameters of the interplanetary magnetic field and solar wind into the training data. We also estimate the effect of adding feature transformations such as adding time-series differences. Forecasting the Kp index is of great practical importance, since strong geomagnetic disturbances lead to undesirable effects, such as the occurrence of geomagnetically induced currents (the strength of which during magnetic storms can reach tens of amperes) in long conductors with low resistance - communication and power lines, pipelines, railways; failures in radio communication systems and satellite navigation systems. Given the 3-hour frequency of the Kp-index, the task is formulated as forecasting next 8 values of the Kp index every hour. We use gradient boosting and perceptron type neural networks, which showed best performance in this task in our previous studies. Previously [1] we used only hourly frequency data available from the ACE Science Center [2] and from other sources (e.g. Dst index from [3]). This is a common approach, because it is relatively simple, it looks consistent, and it does not require a lot of computational resources. However, our analysis, as well as the underlying physics of geomagnetic processes, suggested that higher-frequency data (especially for the Bz component) could serve as a good predictor for geomagnetic disturbances. The ACE Science Center provides historical data on the IMF and SW parameters with a 5-minute frequency (updated daily) [ссылка], | ||
+ | This study has been performed within the framework of the state assignment of M.V.Lomonosov Moscow State University. \\ | ||
+ | [1] Gadzhiev I.M., Barinov O.G., Dolenko S.A., Myagkova I.N., Comparative Analysis of the Procedures to Forecast the Kp Geomagnetic Index by Machine Learning, Moscow University Physics Bulletin 79(2), P. 854-865, [[http:// | ||
+ | [2] ACE Science Center https:// | ||
+ | [3] World Data Center for Geomagnetism, | ||
+ | [4] National Oceanic and Atmospheric Administration, | ||
+ | |||
+ | ==== 96. Машинное обучение для статистической детализации характеристик пространственного распределения осадков в Московском регионе ==== | ||
+ | |||
+ | // | ||
+ | (1) Московский государственный университет имени М.В. Ломоносова, | ||
+ | |||
+ | В связи с наблюдаемыми изменениями климата учащающиеся экстремальные осадки оказывают влияние на различные регионы, | ||
+ | В данной работе методы машинного обучения в статистической детализации используются для получения характеристик пространственного распределения осадков (максимального значения, | ||
+ | Работа выполнена при поддержке Некоммерческого фонда содействия развитию науки и образования «ИНТЕЛЛЕКТ». | ||
+ | |||
+ | ---- | ||
+ | |||
+ | 16 | ||
+ | ===== Section 3. Machine Learning in Natural Sciences ===== | ||
+ | |||
+ | |||
+ | ==== 37. Neutron spectrum unfolding with deep learning models for tabular data ==== | ||
+ | |||
+ | //**Chizhov Konstantin Alekseevich** (1,2), Bely Artyom Alekseevich (2) \\ (1) Joint Institute for Nuclear Research, Laboratory of Information Technologies named after. M.G. Meshcheryakov, | ||
+ | |||
+ | Estimation of the effective dose and unfolding the spectrum of neutron radiation at nuclear power facilities and charged particle accelerators is complicated by the lack of direct methods for detecting neutrons and the need to register secondary particles. The main difficulties are related to the wide energy range of neutrons from 1 meV to several hundred MeV, complex dependence of the neutron interaction cross section on energy. One of the main devices used for neutron spectrometry is the Bonner multi-sphere spectrometer (BSS). The measurement results and the desired spectrum, discretized on the energy grid (or decomposed into basis functions) are tabular data. However, due to the limited set of moderator spheres and correlations in its response functions, the number of input features is limited. In this paper, it is proposed to transform the original scalar continuous features into vectors. And then unfold the spectra for the transformed features using deep learning models included in the Mambular framework: a sequential model from Mamba architecture blocks based on autoregressive state-space models; a model using transform encoders (FT-Transformer), | ||
+ | |||
+ | The research was carried out within the framework of the state assignment of the Ministry of Science and Higher Education of the Russian Federation (topic No. 124112200072-2). | ||
+ | |||
+ | |||
+ | |||
+ | ==== 49. Применение концепции переноса обучения для градиентного бустинга при решении обратных задач разведочной геофизики ==== | ||
+ | |||
+ | // | ||
+ | |||
+ | Данное исследование посвящено изучению эффективности методов переноса обучения, | ||
+ | Исследование выполнено за счёт гранта Российского научного фонда № 24-11-00266. | ||
+ | ==== 48. Анализ стратегий обучения FBPINNs ==== | ||
+ | |||
+ | // | ||
+ | |||
+ | Физически-информированные нейронные сети (PINNs) активно применяются для решения задач механики, | ||
+ | |||
+ | ==== 62. The creation of reasonable robot control behavior in the form of executable code ==== | ||
+ | |||
+ | // | ||
+ | |||
+ | A system for robot control behavior based on Large Language Models using Prompt Engineering methods and answer generation in the form of executable program blocks is presented. The method employed is to convert human instructions into code snippets taking into account information from the robot' | ||
+ | |||
+ | ==== 63. Применение сетей Колмогорова-Арнольда для решения обратной задачи спектроскопии при создании мультимодального наносенсора ионов металлов на основе углеродных точек ==== | ||
+ | |||
+ | // | ||
+ | |||
+ | Разработка мультимодального флуоресцентного наносенсора на основе углеродных точек (УТ) направлена на создание перспективного инструмента для анализа жидких сред и количественной оценки примесей металлов [1]. Одним из ключевых этапов является создание модели, | ||
+ | Исследование выполнено за счет гранта Российского научного фонда № 22-12-00138, | ||
+ | |||
+ | [1] Sarmanova, O.E., Laptinskiy, K.A., Burikov, S.A., Chugreeva, G.N., Dolenko, T.A.: Implementing neural network approach to create carbon-based optical nanosensor of heavy metal ions in liquid media. Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy 286, 122003 (2023). \\ | ||
+ | [2] Liu, Z. et al.: KAN: Kolmogorov-Arnold Networks. arXiv: | ||
+ | [3] Wang, Y. et al.: Kolmogorov–Arnold-Informed neural network: A physics-informed deep learning framework for solving PDEs based on Kolmogorov–Arnold Networks. arXiv: | ||
+ | |||
+ | |||
+ | ==== 64. Optimization of IRT-T research reactor fuel loading pattern by genetic algorithm ==== | ||
+ | |||
+ | //**N.V. Smolnikov** (1), D.V. Pasko (1), M.N. Anikin (1), I.I. Lebedev (1), A.G. Naimushin (1) \\ National Research Tomsk Polytechnic University// | ||
+ | |||
+ | Research Nuclear Reactors (RNR) are powerful sources of neutron and gamma radiation with large number of beams and experimental channels that can be used to produce techninal and medical isotopes, conduct researches in solid physics, neutron scattering and othr fields. Most of RNRs operate in Partial Refueling Mode , where only burnt fuel assemblies are replaced during refueling. This leads to power density increase in a localazied sectors (fuel assemblies), | ||
+ | |||
+ | ==== 59. Камни: Коллективная игра между агентами разнообразных типов, разработанная для изучения взаимодействия человека и искусственного интеллекта в многоагентной среде ==== | ||
+ | |||
+ | // | ||
+ | (1) Физический факультет, | ||
+ | (2) Научно-исследоваельский институт ядерной физики имени Д.В. Скобельцына, | ||
+ | (3) Национальный исследовательский ядерный университет «МИФИ» // | ||
+ | |||
+ | В работе представлена агент-ориентированная игровая система “Камни”, | ||
+ | ==== 70. Построение нейродифференциальных уравнений с применением методов обратных задач динамики ==== | ||
+ | |||
+ | // | ||
+ | |||
+ | В обратных задачах динамики (inverse problems of dynamics) необходимо определить неизвестные силы или параметры динамической системы по известным характеристикам ее движения, | ||
+ | |||
+ | ==== 74. СОЗДАНИЕ ДИНАМИЧЕСКОГО КОГНОВИЗОРА – РАСПОЗНАВАНИЕ КОГНИТИВНЫХ СОСТОЯНИЙ С ПОМОЩЬЮ МЕТОДОВ ГЛУБОКОГО ОБУЧЕНИЯ ==== | ||
+ | |||
+ | // | ||
+ | |||
+ | Данная работа является продолжением исследования, | ||
+ | |||
+ | ==== 78. Analysis of PINN Training Strategies for Heat Conduction Problems ==== | ||
+ | |||
+ | //**Tarasov A.A.**(1), Stepanova M.M.(1), Orlov S.E.(1) \\ (1) Saint-Petersburg State University// | ||
+ | |||
+ | In recent years, physics-informed neural networks (PINNs) have been increasingly employed to solve applied problems in mathematical physics. PINNs incorporate equations and boundary conditions directly into the model architecture through automatic differentiation. This approach allows solving differential equations without explicit spatiotemporal discretization, | ||
+ | |||
+ | ==== 66. Comparison of Data Generation Methods for Spectral Analysis Using Variational Autoencoders ==== | ||
+ | |||
+ | // | ||
+ | |||
+ | This study explores application of data representativity enhancement using variational autoencoders (VAEs) to the inverse problem of Raman spectroscopy of multicomponent aqueous solutions of inorganic salts. By extending our earlier work on optical absorption spectroscopy to Raman scattering, we assess the transferability of VAE-based dataset expansion methods used to solve inverse problems in spectroscopy across spectroscopic techniques. The objective of the considered spectroscopic studies is to determine the concentrations of various ions in multi-component aqueous solutions based on spectral information. Unlike other spectroscopic techniques such as infrared or optical absorption spectroscopy, | ||
+ | This study has been performed at the expense of the grant of the Russian Science Foundation no. 24-11-00266. | ||
+ | |||
+ | ==== 87. Применение переноса обучения сверточной нейронной сети для повышения точности решения обратной задачи фотолюминесцентной наносенсорики ==== | ||
+ | |||
+ | //**Г. Чугреева**(1), | ||
+ | |||
+ | Углеродные точки (УТ) – класс углеродных наночастиц, | ||
+ | Исследование выполнено за счёт гранта Российского научного фонда № 22-12-00138, | ||
+ | [1] Vervald A.M., Laptinskiy K.A., Chugreeva G.N., Burikov S.A., Dolenko T.A.. (2023) Quenching of Photoluminescence of Carbon Dots by Metal Cations in Water: Estimation of Contributions of Different Mechanisms. J. Phys. Chem. C (Vol. 127, pp. 21617-21628). \\ | ||
+ | [2] Wibrianto, A., Khairunisa, S. Q., Sakti, S. C. W., Ni’mah, Y. L., Purwanto, B., & Fahmi, M. Z. (2021). Comparison of the effects of synthesis methods of B, N, S, and P-doped carbon dots with high photoluminescence properties on HeLa tumor cells. RSC Advances (Vol. 11, Issue 2, pp. 1098–1108). \\ | ||
+ | [3] Chugreeva, G. N., Laptinskiy, K. A., Plastinin, I. V., Sarmanova, O. E., & Dolenko, T. A. (2024). Development of a Multimodal Photoluminescent Carbon Nanosensor for Metal Ions in Water Using Artificial Neural Networks. Moscow University Physics Bulletin, 79(S2), S844–S853. | ||
+ | |||
+ | ==== 86. Probabilistic Spiking Neural Network with Correlation-based Memristive Synaptic Updates ==== | ||
+ | |||
+ | //**Dmitry Kunitsyn** (1,2), Alexander Sboev (1,2), Yury Davydov (1), Danila Vlasov (1), Alexey Serenko (1), and Roman Rybka (1,2) \\ (1) National Research Centre “Kurchatov Institute”, | ||
+ | |||
+ | Spiking Neural Networks (SNNs) are a biologically inspired class of neural models that encode information as discrete temporal impulses (spikes). These networks exhibit low latency and reduced power consumption, | ||
+ | |||
+ | ==== 90. Модель машинного обучения для прогнозирования вентиляторных порогов ==== | ||
+ | |||
+ | // | ||
+ | |||
+ | Одной из задач, связанных с определением состояния легочной и сердечно-сосудистой систем, | ||
+ | [1] Beaver WL, Wasserman K, Whipp BJ. A new method for detecting anaerobic threshold by gas exchange. J Appl Physiol, Vol. 60, No. 6, 1985, P. 2020-2027. DOI: 10.1152/ | ||
+ | [2] Mishra, Pradeepta. Practical Explainable AI Using Python: Artificial Intelligence Model Explanations Using Python-based Libraries, Extensions, and Frameworks. Apress Berkeley, CA, 2022. DOI: 10.1007/ | ||
+ | |||
+ | ==== 85. Finding optimal carbon dots synthesis parameters for quantitative analysis of components in multi-component aqueous solutions using machine learning ==== | ||
+ | |||
+ | //**Guskov A.A.** (1, 2), Isaev I.V. (2), Laptinskiy K.A. (2), Dolenko T.A. (1, 2), Dolenko S.A. (2) \\ (1) Faculty of Physics, M.V. Lomonosov Moscow State University, Moscow, Russia, (2) D.V. Skobeltsyn Institute of Nuclear Physics, M.V. Lomonosov Moscow State University, Moscow, Russia// | ||
+ | |||
+ | Optical nanosensors based on carbon dots (CD) introduced into the object of the study are widely used for analyzing the content of multicomponent liquid media. Their applicability stems from the high sensitivity of CD photoluminescence to changes in medium parameters, such as pH and solution temperature, | ||
+ | This study has been performed at the expense of the grant of the Russian Science Foundation no. 22-12-00138-P. | ||
+ | |||
+ | ==== 91. Classifying Russian speech commands with a hardware-deployable spiking neural network transferred from an artificial neural network ==== | ||
+ | |||
+ | //**Roman Rybka** (1, 2), Alexey Serenko (1), Alexander Naumov (1), Alexander Sboev (1, 2) \\ (1) National research centre Kurchatov Institute, Moscow, Russia (2) National research nuclear university MEPhI, Moscow, Russia// | ||
+ | |||
+ | We present a baseline accuracy of classifying audio recordings of command words in Russian from a recent dataset RuSC using a 7-layer convolutional spiking neural network of Integrate-and-Fire neurons. The network is obtained by transferring weights from a trained network of ReLU neurons of same topology, and then by adjusting neuron thresholds using a same-topology network with the ClipFloor activation function. In order to make the network prospectively deployable to neuromorphic processors, its synaptic weights are quantized to 8-bit integer. When the duration of presenting one input sample is 200 time steps of spiking network, the resulting performance is the f1-micro of 98%, which is just 1% lower than originally reported on that dataset with artificial neural networks. This result might be a starting point against which further spiking network solutions for keyword spotting in Russian could be compared. | ||
+ | ---- | ||
+ | |||
+ | 16 |