1. The time evolution of the surface detector of the Pierre Auger ObservatoryOrazio Zapparrata, Andrej Filipčič, Jon Paul Lundquist, Shima Ujjani Shivashankara, Samo Stanič, Serguei Vorobiov, Danilo Zavrtanik, Marko Zavrtanik, 2023, objavljeni znanstveni prispevek na konferenci Opis: The surface detector array of the Pierre Auger Observatory, consisting of 1660 water Cherenkov tanks, has been in operation for nearly 20 years. During this long period of data acquisition, ageing
effects in the detector response have been observed. The temporal evolution of the signals recorded by the surface detector is mostly compensated by continuous calibration with atmospheric muons;
however, effects persist in the signal rise time and in high-level data analysis using neural networks. We have implemented a detailed description of the time evolution of the detector response and of
the uptimes of individual stations in GEANT4-based detector simulations. These new simulations reproduce the observed time dependencies in the data. Using air-shower simulations that take
into account the evolution of individual stations, we show that the reconstructed energy is stable at the sub-percent level, and its resolution is affected by less than 5% in 15 years. For a few
specific stations, the collected light produced by muons has decreased to the point where it is difficult to distinguish it from the electromagnetic background in the calibration histograms. The
upgrade of the Observatory with scintillator detectors mitigates this problem: by requiring a coincidence between the water-Cherenkov and scintillator detectors, we can enhance the muon relative contribution to the calibration histogram. We present the impact and performance of this coincidence calibration method. Ključne besede: surface detector, Pierre Auger Observatory, neural networks, air-shower simulations Objavljeno v RUNG: 22.01.2024; Ogledov: 650; Prenosov: 4 Celotno besedilo (743,29 KB) Gradivo ima več datotek! Več... |
2. Mass composition of Telescope Array's surface detectors events using deep learningI. Kharuk, Jon Paul Lundquist, 2022, objavljeni znanstveni prispevek na konferenci Opis: We report on an improvement of deep learning techniques used for identifying primary particles of atmospheric air showers. The progress was achieved by using two neural networks. The first works as a classifier for individual events, while the second predicts fractions of elements in an ensemble of events based on the inference of the first network. For a fixed hadronic model, this approach yields an accuracy of 90% in identifying fractions of elements in an ensemble of events. Ključne besede: Telescope Array, indirect detection, ground array, surface detection, ultra-high energy, cosmic rays, composition, deep learning, machine learning, neural networks Objavljeno v RUNG: 29.09.2023; Ogledov: 947; Prenosov: 4 Celotno besedilo (788,87 KB) Gradivo ima več datotek! Več... |
3. Application of pattern spectra and convolutional neural networks to the analysis of simulated Cherenkov Telescope Array dataJann Aschersleben, Saptashwa Bhattacharyya, Barbara MARČUN, Judit Pérez Romero, Samo Stanič, Veronika Vodeb, Serguei Vorobiov, Gabrijela Zaharijas, Marko Zavrtanik, Danilo Zavrtanik, Miha Živec, 2021, objavljeni znanstveni prispevek na konferenci Opis: The Cherenkov Telescope Array (CTA) will be the next generation gamma-ray observatory and will be the major global instrument for very-high-energy astronomy over the next decade, o˙ering 5 − 10 × better flux sensitivity than current generation gamma-ray telescopes. Each telescope will provide a snapshot of gamma-ray induced particle showers by capturing the induced Cherenkov emission at ground level. The simulation of such events provides images that can be used as training data for convolutional neural networks (CNNs) to determine the energy of the initial gamma rays. Compared to other state-of-the-art algorithms, analyses based on CNNs promise to further enhance the performance to be achieved by CTA.
Pattern spectra are commonly used tools for image classification and provide the distributions of the shapes and sizes of various objects comprising an image. The use of relatively shallow CNNs on pattern spectra would automatically select relevant combinations of features within an image, taking advantage of the 2D nature of pattern spectra. In this work, we generate pattern spectra from simulated gamma-ray events instead of using the raw images themselves in order to train our CNN for energy reconstruction. This is di˙erent from other relevant learning and feature selection methods that have been tried in the past. Thereby, we aim to obtain a significantly faster and less computationally intensive algorithm, with minimal loss of performance. Ključne besede: Cherenkov Telescope Array, very-high-energy astronomy, convolutional neural networks Objavljeno v RUNG: 18.09.2023; Ogledov: 698; Prenosov: 4 Celotno besedilo (1,24 MB) Gradivo ima več datotek! Več... |
4. Multiple allocation p-hub location problem for content placement in VoD services: a differential evolution based approachSoumen Atta, Goutam Sen, 2020, izvirni znanstveni članek Opis: In video-on-demand (VoD) services, large volumes of digital data are kept at hubs which are spatially distributed over large geographic areas and users are connected to these hubs based on their demands. In this article, we consider a large database of video files, that are pre-partitioned to multiple segments based on the demand patterns of users. These segments are restricted to be located only in hubs. Here, users are allowed to be allocated to multiple hubs and all hubs are assumed to be connected with each other. We jointly decide the location of hubs, the placement of segments to these hubs and then the assignment of users to these hubs as per their demand patterns and finally, we find the optimal paths to route the demands of users for different segments having the objective of minimizing the total routing cost. In this article, a differential evolution (DE) based method is proposed to solve the problem. The proposed DE-based method utilizes an efficient function to evaluate the objective value of a candidate solution to the proposed problem. It also incorporates two problem-specific solution refinement techniques for faster convergence. Instances of the problem are generated from the real world movie database and the proposed method is applied to these instances and the performance is evaluated against the benchmark results obtained from CPLEX. Ključne besede: Video-on-demand (VoD) services, Content distribution networks, Database segment location, Hub location, Multiple hub allocation, Differential evolution (DE), IBM ILOG CPLEX Objavljeno v RUNG: 17.04.2023; Ogledov: 1134; Prenosov: 0 Gradivo ima več datotek! Več... |
5. |
6. |
7. |
8. Comparative analysis of epidemiological models for COVID-19 pandemic predictionsRajan Gupta, Gaurav Pandey, Saibal K. Pal, 2021, izvirni znanstveni članek Opis: Epidemiological modeling is an important problem around the world. This research presents COVID-19 analysis to understand which model works better for different regions. A comparative analysis of three growth curve fitting models (Gompertz, Logistic, and Exponential), two mathematical models (SEIR and IDEA), two forecasting models (Holt’s exponential and ARIMA), and four machine/deep learning models (Neural Network, LSTM Networks, GANs, and Random Forest) using three evaluation criteria on ten prominent regions around the world from North America, South America, Europe, and Asia has been presented. The minimum and median values for RMSE were 1.8 and 5372.9; the values for the mean absolute percentage error were 0.005
and 6.63; and the values for AIC were 87.07 and 613.3, respectively, from a total of 125 experiments across 10 regions. The growth curve fitting models worked well where flattening of the cases has started. Based on region’s growth curve, a relevant model from the list can be used for predicting the number of infected cases for COVID-19. Some other models used in forecasting the number of cases have been added in the future work section, which can help researchers to forecast the number of cases in different regions of the world. Ključne besede: epidemic modeling, machine learning, neural networks, pandemic forecasting, time-series forecasting Objavljeno v RUNG: 15.07.2021; Ogledov: 2437; Prenosov: 33 Povezava na celotno besedilo Gradivo ima več datotek! Več... |
9. Dealing with network changes in cellular fingerprint positioning systemsAndrea Viel, Paolo Gallo, Angelo Montanari, Donatella Gubiani, Andrea Dalla Torre, Federico Pittino, Chris Marshall, 2017, objavljeni znanstveni prispevek na konferenci Opis: Besides being a fundamental infrastructure for communication, cellular networks are exploited for positioning through signal fingerprinting. Maintaining the fingerprint database consistent and up-to-date is a challenging task in many fingerprint positioning systems, e.g., in those populated by a crowd-sourcing effort. To this end, detecting and tracking the changes in the configurations of cellular networks over time is recognized as a relevant problem. In this paper, we show that to cope with this problem we can successfully exploit information provided by Timing Advance (TA). As a by-product, we prove that TA can improve the fingerprint candidate selection phase, reducing the number of fingerprints to provide as input to positioning algorithms. The effectiveness of the proposed improvements has been tested on a fingerprint positioning system with a large fingerprint dataset collected over a period of 2 years. Ključne besede: fingerprint positioning systems, cellular communication networks, network changes Objavljeno v RUNG: 13.06.2018; Ogledov: 4079; Prenosov: 0 Gradivo ima več datotek! Več... |
10. Weak forms of shadowing in topological dynamicsSergey Kryzhevich, Danila Cherkashin, 2017, izvirni znanstveni članek Opis: We consider continuous maps of compact metric spaces. It is proved that every pseudotrajectory with sufficiently small errors contains a subsequence of positive density that is point-wise close to a subsequence of an exact trajectory with the same indices. Also, we study homeomorphisms such that any pseudotrajectory can be shadowed by a finite number of exact orbits. In terms of numerical methods this property (we call it multishadowing) implies possibility to calculate minimal points of the dynamical system.
We prove that for the non-wandering case multishadowing is equivalent to density of minimal points. Moreover, it is equivalent to existence of a family of $\varepsilon$-networks ($\varepsilon > 0$) whose iterations are also $\varepsilon$-networks. Relations between multishadowing and some ergodic and topological properties of dynamical systems are discussed. Ključne besede: Topological dynamics, minimal points, invariant measure, shadowing, chain recurrence, $\varepsilon$-networks, syndetic sets Objavljeno v RUNG: 27.07.2017; Ogledov: 4339; Prenosov: 0 Gradivo ima več datotek! Več... |