{"title":"An Overview of Machine Learning and Deep Learning Applications in Earth Sciences in 2024: Achievements and Perspectives","authors":"M. A. Krinitskiy","doi":"10.3103/S0027134924702217","DOIUrl":"10.3103/S0027134924702217","url":null,"abstract":"<p>Machine learning (ML) and deep learning (DL) methods are extensively applied in various fields of Earth sciences, such as oceanography, meteorology, and climatology. These statistical approaches enable efficient processing of large volumes of data, uncovering hidden patterns, reducing or assessing uncertainty in climate and weather forecasts, automating monitoring, and accelerating analytical research. Among most successful examples, one may mention remote sensing data analysis, geophysical processes modeling, approximating unknown physical parameters, and solving statistical weather and climate forecasting problems. However, there are certain challenges, such as the need for large data volumes, computational demands and technical issues of the data science approach, and ensuring the physical plausibility of results. In the future, the development of hybrid models that combine physical and statistical methods is anticipated, as well as improvements in the interpretability of ML and DL models. In this overview, we will examine current achievements in the application of ML and DL in the study of the ocean, atmosphere, and climate, and we will discuss the challenges and prospects for their further development. This overview places particular emphasis on the progress made in the Russian Federation scientific community regarding the application of ML, DL, and AI within Earth sciences, highlighting both its accomplishments and the challenges it faces in the global research landscape.</p>","PeriodicalId":711,"journal":{"name":"Moscow University Physics Bulletin","volume":"79 2 supplement","pages":"S739 - S749"},"PeriodicalIF":0.4,"publicationDate":"2025-03-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143668290","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
G. N. Chugreeva, K. A. Laptinskiy, I. V. Plastinin, O. E. Sarmanova, T. A. Dolenko
{"title":"Development of a Multimodal Photoluminescent Carbon Nanosensor for Metal Ions in Water Using Artificial Neural Networks","authors":"G. N. Chugreeva, K. A. Laptinskiy, I. V. Plastinin, O. E. Sarmanova, T. A. Dolenko","doi":"10.3103/S0027134924702308","DOIUrl":"10.3103/S0027134924702308","url":null,"abstract":"<p>The study presents a novel approach to the development of a photoluminescent multimodal carbon dots-based nanosensor for determining the concentration of Cu<span>({}^{2+})</span>, Ni<span>({}^{2+})</span>, Co<span>({}^{2+})</span>, Al<span>({}^{3+})</span>, Cr<span>({}^{3+})</span> and NO<span>({}^{-}_{3})</span> ions in water using machine learning methods. The results show that it is possible to determine the type and concentration of each of these ions in multicomponent aqueous media from the photoluminescence spectra of carbon dots. The mean absolute error of simultaneous determination of Cu<span>({}^{2+})</span>, Ni<span>({}^{2+})</span>, Co<span>({}^{2+})</span>, Al<span>({}^{3+})</span>, Cr<span>({}^{3+})</span> cations and NO<span>({}^{-}_{3})</span> anion concentration was 0.85, 0.97, 0.67, 0.81, 0.26, and 2.03 mM, respectively. The accuracy of the developed nanosensor fully meets the requirements for wastewater and process water composition control. The developed nanosensor can not only simultaneously determine the concentration of each of the 6 ions, but also provides real-time remote determination of ion concentrations.</p>","PeriodicalId":711,"journal":{"name":"Moscow University Physics Bulletin","volume":"79 2 supplement","pages":"S844 - S853"},"PeriodicalIF":0.4,"publicationDate":"2025-03-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143667886","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Engineering Point Defects in MoS({}_{textbf{2}}) for Tailored Material Properties Using Large Language Models","authors":"Abdalaziz Al-Maeeni, Denis Derkach, Andrey Ustyuzhanin","doi":"10.3103/S002713492470228X","DOIUrl":"10.3103/S002713492470228X","url":null,"abstract":"<p>The tunability of physical properties in transition metal dichalcogenides (TMDCs) through point defect engineering offers significant potential for the development of next-generation optoelectronic and high-tech applications. Building upon prior work on machine learning-driven material design, this study focuses on the systematic introduction and manipulation of point defects in MoS<span>({}_{2})</span> to tailor their properties. Leveraging a comprehensive dataset generated via density functional theory (DFT) calculations, we explore the effects of various defect types and concentrations on the material characteristics of TMDCs. Our methodology integrates the use of pretrained large language models to generate defect configurations, enabling efficient predictions of defect-induced property modifications. This research differs from traditional methods of material generation and discovery by utilizing the latest advances in transformer model architecture, which have proven to be efficient and accurate discrete predictors. In contrast to high-throughput methods where configurations are generated randomly and then screened based on their physical properties, our approach not only enhances the understanding of defect-property relationships in TMDCs but also provides a robust framework for designing materials with bespoke properties. This facilitates the advancement of materials science and technology.</p>","PeriodicalId":711,"journal":{"name":"Moscow University Physics Bulletin","volume":"79 2 supplement","pages":"S818 - S827"},"PeriodicalIF":0.4,"publicationDate":"2025-03-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143668396","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
M. I. Varentsov, M. A. Krinitskiy, V. M. Stepanenko
{"title":"Approximation of Spatial and Temporal Variability of the Urban Heat Island in Moscow Using Machine Learning","authors":"M. I. Varentsov, M. A. Krinitskiy, V. M. Stepanenko","doi":"10.3103/S0027134924702254","DOIUrl":"10.3103/S0027134924702254","url":null,"abstract":"<p>This study is devoted to the application of machine learning (ML) methods for statistical approximation of the urban-induced temperature anomaly, known as the urban heat island (UHI), and its spatiotemporal dynamics, using the example of the Moscow megacity. This task is considered as part of a more general problem of statistical downscaling of meteorological fields for urban conditions. Therefore, we aim to approximate a high-resolution field of urban temperature anomalies based on predictors characterizing low-resolution meteorological data and high-resolution surface properties. As the input data for training ML models, we use the results of high-resolution hydrodynamic simulations of the meteorological regime in the Moscow region conducted with the COSMO regional atmospheric model coupled with the TERRA_URB urban canopy parameterization. For the ML model, we use the gradient boosting method implemented by the CatBoost algorithm with GPU support. To account for nonlocal dependences between UHI and surface properties, we use an original quasi-local approach to define the feature vectors. This approach consists of using data localized at individual points (nodes of the computational mesh of the COSMO model) as feature descriptions and generating additional features based on the predictors’ values for neighboring points using different types of convolution filters. As such filters, we use a moving average with a circular kernel of different radii and more advanced self-adjusting directional filters formed by taking into account large-scale data on wind speed and direction. We show that such nonlocal features are important for correctly reproducing the key patterns of the UHI spatial structure, in particular the smoother structure of seasonally-averaged temperature anomalies in comparison to surface properties, and the shift of temperature anomalies to the leeward side of the city for specific cases with different wind directions.</p>","PeriodicalId":711,"journal":{"name":"Moscow University Physics Bulletin","volume":"79 2 supplement","pages":"S784 - S797"},"PeriodicalIF":0.4,"publicationDate":"2025-03-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143668398","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
I. M. Gadzhiev, O. G. Barinov, S. A. Dolenko, I. N. Myagkova
{"title":"Comparative Analysis of the Procedures to Forecast the Kp Geomagnetic Index by Machine Learning","authors":"I. M. Gadzhiev, O. G. Barinov, S. A. Dolenko, I. N. Myagkova","doi":"10.3103/S002713492470231X","DOIUrl":"10.3103/S002713492470231X","url":null,"abstract":"<p>Geomagnetic disturbances are one of the most important factors in space weather, the role of which will increase with the development of the space industry and the global digital industry, both on Earth and in near-Earth space. Geomagnetic activity is usually characterized by special indices. One of the most common geomagnetic indices is the Kp index, first introduced by Julius Bartels in 1939. In this study, we explore the possibility of predicting the following Kp index values during the next day using machine learning models based on the hourly values of the parameters of solar wind and interplanetary magnetic field, and of the hourly Dst index. We use such ML models as linear regression, gradient boosting and multilayer perceptrons. We test to what extent the use of history of time series improves the performance of ML models. We draw conclusions about the optimal procedure of creating and applying of a machine learning model to solve the Kp index forecasting problem. The best results by most of the quality metrics were demonstrated by CatBoost and perceptron with two hidden layers. The most significant input features detected were preceding values of the Kp index itself, solar wind velocity and density, modulus and z-component of the interplanetary magnetic field.</p>","PeriodicalId":711,"journal":{"name":"Moscow University Physics Bulletin","volume":"79 2 supplement","pages":"S854 - S865"},"PeriodicalIF":0.4,"publicationDate":"2025-03-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143668138","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Multidimensional Global Optimization of Detector Systems Using the Example of Muon Shield in the SHiP Experiment","authors":"E. O. Kurbatov, F. D. Ratnikov, E. D. Ursov","doi":"10.3103/S0027134924702151","DOIUrl":"10.3103/S0027134924702151","url":null,"abstract":"<p>SHiP (Search for Hidden Particles) is a new general-purpose experiment at the SPS ring at CERN, aimed at searching for hidden particles proposed by numerous theories beyond the Standard Model. An important element of the experiment is muon shield. On one hand, it must provide good background suppression, and on the other hand, it should not be too heavy. This work presents the results of obtaining muon shield configurations using Bayesian optimization with several types of surrogates. This allowed for effective global multidimensional optimization in a 42-dimensional space and reduced the muon flux by 2.5 times while maintaining the original mass of the shield.</p>","PeriodicalId":711,"journal":{"name":"Moscow University Physics Bulletin","volume":"79 2 supplement","pages":"S700 - S705"},"PeriodicalIF":0.4,"publicationDate":"2025-03-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143668049","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
A. N. Balandina, B. V. Gruzdev, N. A. Savelev, Y. S. Budakyan, S. I. Kisil, A. R. Bogdanov, E. A. Grachev
{"title":"A Transformer Architecture for Risk Analysis of Group Effects of Food Nutrients","authors":"A. N. Balandina, B. V. Gruzdev, N. A. Savelev, Y. S. Budakyan, S. I. Kisil, A. R. Bogdanov, E. A. Grachev","doi":"10.3103/S0027134924702291","DOIUrl":"10.3103/S0027134924702291","url":null,"abstract":"<p>In medicine, context is crucial for accurate patient diagnosis, as the same indicator can have different implications based on its setting. Transformer architecture models have not yet been applied to analyze nutritional data in patient histories. These models offer significant advantages for biomedical analysis, such as considering global context, interpreting attention weights, and generating informative input vectors. The attention mechanism’s ability to uncover multifactorial relationships can help physicians save time and concentrate on specific patterns identified by the neural network. This study adapted the encoder transformer for tabular data, applying it to classify metabolic disorders in patient history. Research into applying transformer architecture to tabular and dietary data shows great promise, yielding results that align with established medical findings while introducing innovative methods for utilizing attention and vectorizing this data.</p>","PeriodicalId":711,"journal":{"name":"Moscow University Physics Bulletin","volume":"79 2 supplement","pages":"S828 - S843"},"PeriodicalIF":0.4,"publicationDate":"2025-03-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143668369","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
A. I. Suslov, M. A. Krinitskiy, C. Staquet, E. Le Boudec
{"title":"Machine Learning Methods for Statistical Prediction of PM2.5 in Urban Agglomerations with Complex Terrain, Using Grenoble As an Example","authors":"A. I. Suslov, M. A. Krinitskiy, C. Staquet, E. Le Boudec","doi":"10.3103/S0027134924702242","DOIUrl":"10.3103/S0027134924702242","url":null,"abstract":"<p>In this study, we propose several methods based on machine learning approaches for predicting air pollution levels in cities located in mountain valleys, with Grenoble (France) as a case study. Pollution forecasting is performed using both regression and classification of exceeding threshold levels. We employ a data-driven approach, utilizing various machine-learning models. Based on historical data from 2012 to 2018, collected at several meteorological stations in the Grenoble Valley, multiple machine learning models were trained to predict the daily average concentrations of fine particulate matter PM10 and PM2.5 three days ahead. Days with high PM concentrations exceeding the threshold values set by the World Health Organization (WHO) are of particular interest in our study. It was found that the presence of local meteorological conditions leads to the formation of temperature inversions, which are statistically associated with air pollution levels in this region. Although local meteorological conditions primarily determine the pollution level, the machine learning models considered in our study can be adapted for other cities in valleys by training them on relevant data.</p>","PeriodicalId":711,"journal":{"name":"Moscow University Physics Bulletin","volume":"79 2 supplement","pages":"S774 - S783"},"PeriodicalIF":0.4,"publicationDate":"2025-03-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143668399","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Pointwise and Complex Quality Metrics in Atmospheric Modeling: Methods and Approaches","authors":"V. Yu. Rezvov, M. A. Krinitskiy, M. A. Borisov","doi":"10.3103/S0027134924702229","DOIUrl":"10.3103/S0027134924702229","url":null,"abstract":"<p>In atmospheric sciences, various quantitative indicators, or metrics, are used to describe the quality of modeling results of various flavors including numerical weather prediction, statistical correction, various downscaling products, etc. Metrics provide the accuracy of reproduction of the processes underlying the models and allow comparison of models by assessing the uncertainty of their results. The key importance of metrics lies in a more thorough study of the advantages and disadvantages of classical approaches and in the development of new, more complex assessment methods. This article presents a classification of the most frequently encountered quality metrics in the scientific literature. In addition to assessing traditional pointwise metrics, complex methods considering various aspects of modeling results and special metrics used in climate studies are described. Among the complex metrics, methods with an emphasis on the spatial structure and heterogeneity of the predicted variable fields and probabilistic methods for verifying ensemble forecasts are distinguished. Special attention in this paper is devoted to the growing popularity of object-oriented metrics and metrics based on rare and extreme events. Climate models are assessed by comparing the results of retrospective modeling with historical data, which complicates the choice of metrics. A variety of climate metrics focusing on specific climate processes or integrating several parameters is described. The need for developing more diverse metrics for effective evaluation of climate models is explored. All metrics considered in this article are supplemented by examples in the scientific literature and assessments of their application to atmospheric research.</p>","PeriodicalId":711,"journal":{"name":"Moscow University Physics Bulletin","volume":"79 2 supplement","pages":"S750 - S764"},"PeriodicalIF":0.4,"publicationDate":"2025-03-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143668289","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
I. S. Lazukhin, M. I. Petrovskiy, I. V. Mashechkin
{"title":"Feature Selection Methods for Deep Learning Models of Soft Sensors in Oil Refining","authors":"I. S. Lazukhin, M. I. Petrovskiy, I. V. Mashechkin","doi":"10.3103/S0027134924702333","DOIUrl":"10.3103/S0027134924702333","url":null,"abstract":"<p>The development of automated control systems results into industrial plants accumulating large amounts of data on the continuous state of technological processes. Multiple physical sensors record the system states at any given time, hence being crucially responsible for controlling the system and maintaining its parameters within hard limits. At the same time, irregularly conducted laboratory measures make up a significant part of the qualitative indicators of such processes, especially in the petrochemical industry. Mathematical models that generalize laboratory measured indicators to match the frequency of physical sensors are called soft sensors. On practice, soft sensors for laboratory data are represented by linear or last-recorded-value models. We investigate the task of analytically obtaining chemical indicators of the technological process in real time based on the values of physical sensors; the study is conducted on a real-world data set. Several problems are covered, including high dimension of the physical inputs compared to the laboratory data volume; scarcity of the laboratory data collected on a daily basis. Authors propose feature selection methods based on PLS regression (hierarchical clustering), Bayes trees, utilize existing graph neural network, as well as compare developed methods with existing popular approaches. Each of the proposed feature selection methods has been adapted to take into account the expert opinion of the industrial plant engineers. Authors investigate developed approaches alongside neural network methods for predicting time series including graph neural networks, fully connected and recurrent networks. The obtained experimental results show the advantage of using proposed feature selection based on PLS and Bayes in ensemble with simple recurrent networks or graph neural networks with preliminary interpolation. Separately, it is worth noting the ambiguity of assessing the developed models quality; authors propose a combined approach that takes into account the adequacy of the model, its correlation with the true laboratory values and averaged errors.</p>","PeriodicalId":711,"journal":{"name":"Moscow University Physics Bulletin","volume":"79 2 supplement","pages":"S872 - S889"},"PeriodicalIF":0.4,"publicationDate":"2025-03-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143667883","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":4,"RegionCategory":"物理与天体物理","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}