The extended Kalman filter (EKF) is a cornerstone of nonlinear state estimation, yet its performance is fundamentally limited by noise-model mismatch and linearization errors. We develop a residual-aware distributionally robust EKF that addresses both challenges within a unified Wasserstein distributionally robust state estimation framework. The key idea is to treat linearization residuals as uncertainty and absorb them into an effective uncertainty model captured by a stage-wise ambiguity set, enabling noise-model mismatch and approximation errors to be handled within a single formulation. This approach yields a computable effective radius along with deterministic upper bounds on the prior and posterior mean-squared errors of the true nonlinear estimation error. The resulting filter admits a tractable semidefinite programming reformulation while preserving the recursive structure of the classical EKF. Simulations on coordinated-turn target tracking and uncertainty-aware robot navigation demonstrate improved estimation accuracy and safety compared to standard EKF baselines under model mismatch and nonlinear effects.
The Rice distribution is applied as a mathematical model in the study of various problems in science and technology. The main task in applications is to estimate the parameters of the Rice distribution from a sample of the measured signal and to separate the parameters of the deterministic signal and noise based on these estimates. Parameter estimation is mainly performed using the maximum likelihood method and the method of moments. However, as is known, both methods often lead to solving systems of equations containing special functions, so additional computational resources are used for the solution. One of the methods to overcome these difficulties is the development of simple, yet sufficiently effective empirical formulas (EFs) that are competitive in accuracy with known algorithms for estimating the parameters of the Rice distribution. This work is devoted to solving this problem.
Atmospheric haze significantly degrades image quality in critical applications such as autonomous driving and surveillance systems, where clear visibility is essential for safety and performance. A fundamental challenge in image dehazing is bridging the gap between the synthetic and real haze domains. Existing deep learning methods show promising results on synthetic datasets, but their effectiveness diminishes dramatically in real-world scenarios. This degradation arises from the diverse and complex characteristics of atmospheric haze, domain shift between synthetic training data and real-world conditions, and the inherent ill-posed nature of the dehazing problem. We present a multiscale variational Bayesian model that integrates a prior on the haze-free image with three complementary components that preserve spatial structure, provide noise robustness, and frequency-domain consistency across different resolutions. This enables the model to dynamically adapt to varying haze distributions unseen during training. Furthermore, we incorporate an adaptive network architecture with input-conditioned convolutions to ensure physical consistency with the atmospheric scattering model, which in turn provides crucial inductive bias for cross-domain generalization. Despite training exclusively on synthetic data, our method demonstrates superior synthetic-to-real generalization capabilities, achieving state-of-the-art performance on multiple real-world benchmarks, while remaining competitive on the synthetic test set. Furthermore, we demonstrate the practical utility of our approach through the evaluation of a downstream task on the KITTI benchmark, showing that our dehazing method significantly improves object detection performance in realistic driving scenarios compared to current state-of-the-art approaches.
The integration of artificial intelligence (AI) into the financial system, especially in the banking sector, has become one of the most important directions of technological progress. The aim of the article is to reveal the specifics of the application of AI in banking services, emphasizing the role of chatbots and virtual assistants in the customer-centric services and risk management system. The article presents the theoretical foundations and main directions of application of AI in the banking system. Four main directions are analyzed: customer-centric solutions, process optimization, banking services market management, and improvement of regulatory mechanisms. The experience of international banks (Bank of America, HSBC, DBS, Armenians banks, and others) indicates that the use of AI contributes to reducing operating costs and accelerating and personalizing customer service. However, the integration of AI also raises challenges related to data privacy, cybersecurity, legislative regulations, and the transformation of professional skills. Special attention is paid to the field of credit scoring, where machine learning methods allow for a more accurate assessment of borrower behavior and reduce financial risks. The relevance of the research is due to the fact that AI is no longer an additional technology for banks but a necessary tool for maintaining competitiveness and sustainable development. Received: 21 September 2025 | Revised: 5 January 2026 | Accepted: 10 March 2026 Conflicts of InterestThe authors declare that they have no conflicts of interest to this work. Data Availability Statement The data that support the findings of this study are openly available in PES at https://doi.org/10.24874/PES06.02.023, reference number [31], in ASPUR at https://doi.org/10.61552/JAI.2024.01.004, reference number [32], in International Accountancy Training Centre at https://doi.org/10.59503/29538009-2024.2.14-121, reference number [33]. Author Contribution Statement Suren H. Parsyan: Conceptualization, Methodology, Investigation, Resources, Writing – original draft, Writing – review & editing, Supervision, Project administration. Frida F. Baharyan: Conceptualization, Investigation, Resources, Data curation, Writing – original draft. Gayane A. Avagyan: Conceptualization, Methodology, Validation, Formal analysis, Resources, Writing – original draft, Writing – review & editing. Sergo A. Episkoposian: Methodology, Validation, Visualization. Vardan S. Aleksanyan: Software, Investigation, Supervision. Ararat Kostanian: Writing – review & editing, Visualization. Lilik M. Beglaryan: Methodology, Visualization, Writing – review & editing.
The influence of a black silicon (b-Si) interlayer on the photovoltaic characteristics of tandem perovskite/silicon cells was investigated by numerical modeling in the SCAPS-1D software environment. It is shown that a 640 nm thick nanotextured b-Si interlayer increases the efficiency of the modeled device from 27.17 to 28.97
