Системні дослідження та інформаційні технології: міжнародний науково-технічний журнал, № 3

Постійне посилання зібрання

Переглянути

Нові надходження

Зараз показуємо 1 - 9 з 9
  • ДокументВідкритий доступ
    Multi-step prediction in linearized latent state spaces for representation learning
    (КПІ ім. Ігоря Сікорського, 2022) Tytarenko, A.
    In this paper, we derive a novel method as a generalization over LCEs such as E2C. The method develops the idea of learning a locally linear state space by adding a multi-step prediction, thus allowing for more explicit control over the curvature. We show that the method outperforms E2C without drastic model changes which come with other works, such as PCC and P3C. We discuss the relation between E2C and the presented method and derive update equations. We provide empirical evidence, which suggests that by considering the multi-step prediction, our method – ms-E2C – allows learning much better latent state spaces in terms of curvature and next state predictability. Finally, we also discuss certain stability challenges we encounter with multi-step predictions and how to mitigate them.
  • ДокументВідкритий доступ
    Cost effective hybrid genetic algorithm for workflow scheduling in cloud
    (КПІ ім. Ігоря Сікорського, 2022) Kumar Bothra, Sandeep; Singhal, Sunita; Goyal, Hemlata
    Cloud computing plays a significant role in everyone’s lifestyle by snugly linking communities, information, and trades across the globe. Due to its NP-hard nature, recognizing the optimal solution for workflow scheduling in the cloud is a challenging area. We proposed a hybrid meta-heuristic cost-effective load-balanced approach to schedule workflow in a heterogeneous environment. Our model is based on a genetic algorithm integrated with predict earliest finish time (PEFT) to minimize makespan. Instead of assigning the task randomly to a virtual machine, we apply a greedy strategy that assigns the task to the lowest-loaded virtual machine. After completing the mutation operation, we verify the dependency constraint instead of each crossover operation, which yields a better outcome. The proposed model incorporates the virtual machine’s performance variance as well as acquisition delay, which concedes the minimum makespan and computing cost. One of the most astounding aspects of our cost-effective hybrid genetic algorithm (CHGA) is its capacity to anticipate by creating an optimistic cost table (OCT) while maintaining quadratic time complexity. Based on the results of our meticulous experiments on some real-world workflow benchmarks and comprehensive analysis of some recently successful scheduling algorithms, we concluded that the performance of our CHGA is melodious. CHGA is 14.58188%, 11.40224%, 11.75306%, and 9.78841% cheaper than standard Ant Colony Optimization (ACO), Particle Swarm Optimization (PSO), Cost Effective Genetic Algorithm(CEGA), and Cost-Effective Loadbalanced Genetic Algorithm (CLGA), respectively.
  • ДокументВідкритий доступ
    Методи та моделі нейромережевої апроксимації градуювальних характеристик NTC-термісторів
    (КПІ ім. Ігоря Сікорського, 2022) Федін, С. С.; Зубрецька, І. С.
    Підтверджено гіпотезу про доцільність застосування RBF-мереж для підвищення точності побудови градуювальних характеристик NTC- термісторів у робочому діапазоні температур без поділу його на піддіапазони. Встановлено, що похибка нейромережевої апроксимації градуювальних харак- теристик NTC-термісторів на основі RBF-мереж не менше ніж у півтора рази нижча за допустиму похибку апроксимації поліноміальної моделі третього по- рядку, яка використовується в програмному забезпеченні сучасних систем збирання та оброблення вимірювальної інформації. Розроблено методику об- роблення вимірювальної інформації з використанням RBF-мереж для автома- тизації процедури побудови індивідуальних градуювальних характеристик і періодичного калібрування NTC-термісторів.
  • ДокументВідкритий доступ
    Resource scheduling in edge computing IoT networks using hybrid deep learning algorithm
    (КПІ ім. Ігоря Сікорського, 2022) Vijayasekaran, G.; Duraipandian, M.
    The proliferation of the Internet of Things (IoT) and wireless sensor networks enhances data communication. The demand for data communication rapidly increases, which calls the emerging edge computing paradigm. Edge computing plays a major role in IoT networks and provides computing resources close to the users. Moving the services from the cloud to users increases the communication, storage, and network features of the users. However, massive IoT networks require a large spectrum of resources for their computations. In order to attain this, resource scheduling algorithms are employed in edge computing. Statistical and machine learning-based resource scheduling algorithms have evolved in the past decade, but the performance can be improved if resource requirements are analyzed further. A deep learning-based resource scheduling in edge computing IoT networks is presented in this research work using deep bidirectional recurrent neural network (BRNN) and convolutional neural network algorithms. Before scheduling, the IoT users are categorized into clusters using a spectral clustering algorithm. The proposed model simulation analysis verifies the performance in terms of delay, response time, execution time, and resource utilization. Existing resource scheduling algorithms like a genetic algorithm (GA), Improved Particle Swarm Optimization (IPSO), and LSTM-based models are compared with the proposed model to validate the superior performances.
  • ДокументВідкритий доступ
    Процес керування захищеністю даних під час віддаленої біометричної автентифікації
    (КПІ ім. Ігоря Сікорського, 2022) Астраханцев, А. А.; Ляшенко, Г. Є.
    Системи віддаленої біометричної автентифікації за останній час набули значного поширення через необхідність користування загальними пристроями та виконання платежів через Інтернет. Оскільки саме біометричні методи більш зручні для користувачів і нині швидко замінюють паролі, то стає актуальним завданням передавання біометричної інформації відкритою мережею без її компрометації. Метою роботи є модернізація системи віддаленої автентифікації для підвищення рівня прихованості і захищеності біометричних даних користувача. Запропоновано застосування найкращих за критерієм захищеності методів формування біометричного шаблону, методів мережевої стеганографії для підвищення прихованості та впровадження інтелектуальної системи прийняття рішень. Такі вдосконалення дозволять підвищити захищеність і прихованість даних для процесу віддаленої автентифікації.
  • ДокументВідкритий доступ
    The use of environmental decision support systems for modeling of atmospheric pollution following the chemical accidents
    (КПІ ім. Ігоря Сікорського, 2022) Kovalets, I. V.; Bespalov, V. P.; Maistrenko, S. Ya.; Udovenko, O. I.
    We studied the possibility of the combined application of screening models to assess the characteristics of sources in accidents at storage facilities for hazardous substances with complex models of atmospheric transport as part of modern decision support systems to calculate air pollution in a wide range of spatial and temporal scales. The evaporation time following an emergency spill, estimated by screening models, is used to set the emission intensity and calculate the atmospheric transport by the RODOS nuclear emergency response system. For the accident in Chernihiv on March 23, 2022, it was estimated that the maximum permissible concentration of ammonia 0.2 mg/m3 was exceeded at distances up to 75 km from the source. The dependence of the calculated maximum concentrations on time is close to asymptote 4.5 max c t up to 15 h after emission, which is consistent with the asymptote t 2 / 3 for the time dependence of the sizes of puffs following turbulent dispersion of instantaneous releases.
  • ДокументВідкритий доступ
    Сombined control of multirate impulse processes in a cognitive map of COVID-19 morbidity
    (КПІ ім. Ігоря Сікорського, 2022) Romanenko, V.; Miliavskyi, Y.
    In this article, a cognitive map (CM) of COVID-19 morbidity in a given region was built. A general linear impulse process (IP) model in the CM was developed and measured, and unmeasured CM node coordinates were defined. The general IP model was decomposed into interrelated subsystems with measurable and unmeasurable node coordinates. For the subsystem with measurable node coordinates, multirate sampling of coordinates was conducted, resulting in the development of discrete dynamics models for quickly and slowly measured node coordinates. External controls were selected in IP models based on the possible variation of resources of node coordinates and CM weighting coefficients. IP control laws based on the variation of CM nodes and weight were designed. As a result, recurrent procedures for control generation in closed-loop control subsystems with multirate sampling were formulated. Experimental research on the control subsystems was carried out. It confirmed high efficiency for decreasing COVID-19 morbidity.
  • ДокументВідкритий доступ
    Superconducting gravimeters based on advanced nanomaterials and quantum neural network
    (КПІ ім. Ігоря Сікорського, 2022) Yatsenko, V.; Kruchinin, S.; Bidyuk, P.
    The paper is focused on a new concept of a cryogenic-optical sensor intended for use in the space industry, geodynamics, and fundamental experiments. The basis of the sensor is a magnetic suspension with a levitating test body, a highprecision optical recorder of mechanical coordinates of the levitating body, and a signal-processing system. A Michelson-type interferometer with a laser diode and a single-mode optical fiber was used to measure the test body's displacements. The coordination of the laser diode coherence length and the difference in the interferometer optical lengths of the arms made it possible to eliminate coherent noise caused by interference from spurious reflections. The minimum recorded shift of the test body was 0.1 nm. The design of the sensor and the mathematical model of the superconducting suspension dynamics are presented. The results of experimental studies of a magnetic suspension together with an optical interferometric displacement sensor having a subnanometer sensitivity are shown.
  • ДокументВідкритий доступ
    Study of security trends of the global society based on intelligent data analysis
    (КПІ ім. Ігоря Сікорського, 2022) Zgurovsky, M.; Pyshnograiev, I.
    This article is devoted to applying system analysis and data mining methodology to one of the most pressing problems today: studying the security of a global society in a conflicting world. A set of global threats relevant to the first half of the 21st century is considered. These threats have been identified by the United Nations (UN), the World Health Organization (WHO), the World Economic Forum, and other reputable international organizations. As a result of applying the Delphi method to analyze a wide range of threats identified by these organizations, 11 of the most important threats to humanity in the first half of the 21st century were identified. The vulnerabilities of different countries to the impact of the totality of these threats are analyzed. Scenarios for the possible development of a global society during and after the conflict are constructed.