Speech command recognition (SCR) is the task of detecting predefined voice commands in an audio stream. SCR is used widely in voice assistants, augmented reality, and robotics. Nowadays, Google Speech Commands (GSC) is a benchmark dataset for training and testing SCR models for English. In this work, we developed an open-source speech commands dataset for the low-resourced Tatar language. The dataset employs 35 commands in the GSC dataset and contains 3,547 utterances. To prove the efficacy of our dataset, we trained and evaluated the Keyword-MLP model on our dataset. The model achieved an accuracy of 98.28% on the test set. We optimized the model by converting it to the ONNX Runtime format and deployed it on an NVIDIA Jetson Orin NX 16GB. The optimized model showed an average inference time of 0.003 s on the device, making it suitable for real-time IoT and robotics applications. We have made the dataset, source code, and pretrained models publicly available at https://github.com/IS2AI/TatarSCR to promote the development of voice-controlled systems for the Tatar language.

Federated learning (FL) enables multiple parties to collaboratively train machine learning models while preserving data privacy. However, securing communication within FL frameworks remains a significant challenge due to potential vulnerabilities to data breaches and integrity attacks. This paper proposes a novel approach using Dilithium, a robust digital signature framework, to enhance data security in FL. By integrating Dilithium into FL protocols, this study demonstrates enhanced protection against data tampering and unauthorized access, thereby promoting safer and more efficient collaborative model training across distributed networks. Furthermore, our approach incorporates an optimized key distribution mechanism that reduces latency and ensures seamless synchronization among participants. Additionally, the proposed client selection algorithm aims to minimize the time and computational load on the system. Experimental results demonstrate that our system achieves a total processing time of 6.891 seconds, significantly outperforming the 10.24 seconds of normal FL and 12.32 seconds of FL-Dilithium systems on the same computing platforms. Additionally, the proposed model achieves an accuracy of 94%, surpassing the 93% of the normal FL and 92% of the FL-Dilithium models.

The utilization of Unmanned Aerial Vehicle(UAV) swarms has witnessed a remarkable rise across diverse domain, accompanied by an escalating complexity of challenges. This study addresses these challenges by concurrently considering task rescue value of and UAV type, while also introduce ground vehicles to UAVs recharging or supplies supplementation. The primary objective of this paper is to construct a multi-objective multi-task allocation model for UAV swarms, with the aim of minimizing rescue penalties and resource utilization rates. To tackle this model, the study proposes a Multi-Objective Particle Swarm Optimization based on Decomposition Learning (MPSO-DL). The MPSO-DL algorithm incorporates several innovative features. Firstly, it employs an initialization method that combines clustering and greedy selection techniques to enhance the quality of the initial population; secondly, a particle updating strategy based on solution decomposition learning is devised to expedite the convergence rate of the algorithm; Finally, a local search strategy, leveraging a knowledge base, is developed to increase the probability of the population escaping local optima. Through simulation experiments under three different scenarios, the proposed algorithm is rigorously evaluated. The results demonstrate that the proposed algorithm's ability to yield a set of task allocation strategies characterized by superior convergence and distribution.

Rank metrics quantify machine learning model prediction errors in the setting of binary classification with a given dataset, but imbalances in the class distribution or model predictions can significantly influence the interpretation. We analyze and interpret the area under the receiver operating characteristic curve (auROC) and an area under the precision recall curve, or average precision score (AP), in the presence of two kinds of imbalance: dataset class imbalance and model prediction imbalance. We show that the auROC can have a symmetric but not necessarily uni-modal response to dataset imbalance, but it is insensitive to model prediction imbalance. The AP has an asymmetric response that we show is also not uni-modal. The AP and auROC scores can be challenging to interpret due to these properties. We propose a novel precision recall space, the PR*, and associated average precision score, called AP*, to address the identified shortcomings of the AP and auROC. The proposed AP* metric has a symmetric response to class imbalance, like the auROC, a sensitivity to model prediction imbalance, like the AP, and it adds both a symmetry to model prediction imbalance and unimodal response to class imbalance that makes it more reliable for model selection than the AP or auROC. For model selection, the AP* gives a higher score to the model that tends to make most of its errors in the majority class.

In this paper, we extend our previously proposed Adaptive Continuous Adversarial Training (ACAT) method beyond the problem-space of SPAM filters to encompass the feature-space of Machine Learning (ML)-based Network Intrusion Detection Systems (NIDS). ACAT continuously improves model robustness by incorporating adversarial samples detected from the ML-NIDS input network traffic into the training process. Problem-space evasion attacks rely on inverse-feature mapping, where modifications to real-world objects result in targeted perturbations within the feature vector after feature extraction. Therefore, protecting ML systems against feature-space attacks inherently provides defense against both feature-space and problem-space evasion attacks. Our results demonstrate that ACAT significantly reduces adversarial sample detection time compared to traditional techniques. Moreover, the performance of the ML-based NIDS under attack improved dramatically, with accuracy increasing from 50.92% to over 99% after only three retraining sessions.

An approach for calculation of the magnetic force arising between two electric current-carrying filaments having a circular and closed-curve of arbitrary shape is developed. The developed approach is based on the recently formulated segmentation method applied for the calculation of the mutual induction for a similar filament system. Employing the fact that any curve can be interpolated by a set of line segments with the desired accuracy and deriving the set of formulas for calculating of the magnetic force between a circular filament and line segment, the developed approach was also successfully applied for the estimation of the distribution of magnetic force along the closed-curve in addition to the resulting one. As illustrative examples, the calculation of the magnetic force and its distribution between the circular filament and the following closed-curves such as polygons, circles and a 3D curve was efficiently performed by using the developed approach. Also, the developed method was applied for the calculation of the resultant magnetic force between the rigid bodies including permanent magnets and current-carrying coils. The results of calculation were validated successfully by using FEM method and the analytical formulas available in the literature.

Generative Artificial Intelligence (AI) is revolutionizing educational technology by enabling highly personalized and adaptive learning environments within Intelligent Tutoring Systems (ITS). This report delves into the integration of Generative AI, particularly large language models (LLMs) like GPT-4, into ITS to enhance personalized education through dynamic content generation, real-time feedback, and adaptive learning pathways. We explore key applications such as automated question generation, customized feedback mechanisms, and interactive dialogue systems that respond to individual learner needs. The report also addresses significant challenges, including ensuring pedagogical accuracy, mitigating inherent biases in AI models, and maintaining learner engagement. Future directions highlight the potential advancements in multimodal AI integration, emotional intelligence in tutoring systems, and the ethical implications of AI-driven education. By synthesizing current research and practical implementations, this report underscores the transformative potential of Generative AI in creating more effective, equitable, and engaging educational experiences.

This research investigates predictive maintenance strategies using the AI4I 2020 dataset, a synthetic dataset reflecting real-world industrial environments. The study identifies failure trends, explores their root causes, and evaluates cost impacts associated with different failure types. Advanced analytics reveal heat dissipation failures (HDF) as the most significant contributor to operational disruptions, accounting for over 74% of total maintenance costs. Through a detailed examination of machine operational parameters and failure distribution patterns, this work proposes tailored maintenance solutions, such as enhanced cooling systems, load optimization, and power stabilizers, to mitigate failures effectively. The findings underscore the importance of prioritizing high-impact failures, leveraging predictive maintenance, and adopting proactive strategies to minimize downtime and optimize costs in industrial settings. This research serves as a comprehensive guide for industries aiming to enhance operational reliability and sustainability.

Field Programmable Gate Arrays find extensive use across various domains, ranging from telecommunications to Machine Learning tasks. In the context of space applications, they emerge as a highly favourable choice, owing to their high degree of flexibility and their accessibility in different silicongrade technologies. However, due to the high associated costs of radiation-hardened devices, the prevalence of radiation-tolerant devices is more common in practical deployment scenarios. In this sense, robustness-enhancing techniques can still be applied, thus allowing the employment of these devices in long-duration tasks without the need for a radiation-hardened counterpart. In this paper, we propose a modelling approach for estimating the reliability of a digital design system in space before carrying out fault injection tests and radiation campaigns. For this purpose, we employ a high-level tool called Möbius to obtain a first reliability estimation against Single Event Upsets by means of Fault Tree Models. We also propose a novel methodology for defining which fault mitigation techniques should be applied to each design sub-module, considering the most commonly used in the space environment. In particular, we focus on a specific use case, namely a Soft Graphic Processing Unit IP, implemented on the Xilinx RT XQRKU060 FPGA. We analyze the criticality, power and area impact of the most important design parts of the GPU when implemented in hardware, introducing a classification approach for associating an appropriate fault mitigation technique with each of them. Finally, in the last section of the paper, we hint at future development and conclude the work.

One of the prime problems of computer science and machine learning is to extract information efficiently from large-scale, heterogeneous data. Text data, with its syntax, semantics, and even hidden information content, possesses an exceptional place among the data types in concern. The processing of the text data requires embedding, a method of translating the content of the text to numeric vectors. A correct embedding algorithm is the starting point for obtaining the full information content of the text data. In this work, a new text embedding approach, namely the Guided Transition Probability Matrix (GTPM) model is proposed. The model uses the graph structure of sentences to capture different types of information from text data, such as syntactic, semantic, and hidden content. Using random walks on a weighted word graph, GTPM calculates transition probabilities to derive text embedding vectors. The proposed method is tested with real-world data sets and eight well-known and successful embedding algorithms. GTPM shows significantly better classification performance for binary and multi-class datasets than well-known algorithms. Additionally, the proposed method demonstrates superior robustness, maintaining performance with limited (only 10%) training data, showing an 8% decline compared to 15−20% for baseline methods.

A large-scale IT outage on July 19th, 2024, primarily affecting Windows systems with CrowdStrike software, caused significant disruptions in airlines, finances, and government services. The culprit was a faulty CrowdStrike update, triggering system crashes. We analyze this incident through the lens of the "release first, fix later" approach discussed in the literature, emphasizing technology managers' need for alternative mitigation strategies, including patch management solutions and disaster recovery plans. Additionally, policymakers may explore measures to incentivize thorough testing, while consumers require robust business continuity plans. Vigilance from all parties is crucial in the ever-evolving software landscape.

This article explores the integration of weather balloons within non-terrestrial networks (NTNs) as a pioneering solution for smart cities, addressing both connectivity and security challenges. Weather balloons are proposed as a feasible alternative to satellites, offering seamless connectivity by overcoming terrestrial obstructions and reducing signal latency. Meanwhile, the application of Quantum Key Distribution (QKD), combined with AI optimization, ensures robust data security in high-speed transmission environments. This dual-layer approach forms a comprehensive end-to-end solution, where weather balloons and quantum security work in tandem to provide a resilient, scalable, and secure network infrastructure. The synergy of these technologies is discussed, highlighting their potential to revolutionize smart city infrastructure and set new standards for consumer electronics.

Formulas for both earth return and internal impedances of submarine HVDC cables have been provided in past studies; however, they are based on simplifying hypotheses, e.g. considering 2D geometries, treating armor layers as coaxial tubular conductors, and neglecting the proximity effect. In this paper, a Finite Element analysis investigates the effects of the deviations of realistic geometries and configurations of submarine HVDC cables from idealized ones. In particular, the impact of the twisting of armor wires on internal impedances and the proximity effect in multiple cable configurations on earth return impedance are analyzed.

This comprehensive review critically evaluates the evolving role of hybrid AC/DC power systems in modern energy landscapes, emphasizing their pivotal function in integrating both renewable and conventional power sources. As the integration of renewable resources scales up, optimizing operational efficiency and stability becomes increasingly challenging. This review delves into the advanced optimal power flow (OPF) strategies that not only improve load distribution and energy efficiency but also enhance the reliability and resilience of power grids. By exploring various OPF methodologies, including conventional, probabilistic, and security-constrained OPF, this study underscores the importance of innovative approaches to ensure economic, safe, and reliable grid operations under complex and dynamic energy demands.

One challenge in oceanographic analysis is the need for accurate initial conditions collected from physical buoys. Temporary sensor outages or noisy conditions can hinder the data collection process. Machine learning surrogate models offer short-term coverage during outages. This study presents a methodology for regularizing machine learning models that predict buoy observations by utilizing multiple data sources. A previous work introduced a ratio-coupling hyperparameter to combine numerically modeled data and ocean observations when calculating training loss. However, applying one ratio across all features failed to capture the unique characteristics of different data sources. To overcome this limitation, this work investigates a multiple-hyperparameter loss function to independently manage the contribution of each data source per feature. A bounded random grid search explores the hyperparameter space to find ratios which produce superior results compared to the singleratio approach. Surrogate models are validated at the same 88 fixed locations as the previous paper for a direct comparison. The experimental results suggest that this multi-ratio methodology can offer more reliable forecasts over a 24-hour period by applying the correct weight for each pairing of observed feature and numerical model source.

We consider target parameter estimation for automotive Phase-Modulated Continuous-Wave (PMCW) radar equipped with one-bit modulo analog-to-digital converters (ADCs) with known folding-counts (Mod-ADC-K). One-bit Mod-ADC-K can save energy and overcome the dynamic range problems of one-bit or low-bit conventional ADCs. However, each one-bit Mod-ADC-K introduces an additional folding making it necessary to devise new target parameter estimation algorithms. The Iterative Adaptive Approach (IAA) is a robust nonparametric spectral estimation algorithm with high resolution. We introduce an extended IAA algorithm, referred to as Mod-IAA for one-bit Mod-ADC-K based target parameter estimation. We incorporate the folding-count data generated by Mod-ADC-K into the likelihood function. This likelihood function is then used to construct a proper cost function. Cyclic and Majorization-Minimization (MM) approaches are employed to iteratively optimize this cost function, to estimate the target parameters. Moreover, we employ a relaxation-based maximum likelihood (ML) approach to iteratively refine the results obtained from Mod-IAA for enhanced off-grid target parameter estimation. We also derive the Cramér-Rao bound (CRB) for the target parameter estimation using one-bit Mod-ADC-K based PMCW radar. Furthermore, to determine the number of targets, we derive the Bayesian Information Criterion (BIC) for the one-bit Mod-ADC-K based PMCW radar system. Finally, we provide numerical examples to demonstrate the advantages of using onebit Mod-ADC-K for target parameter estimation using PMCW radar.

This study investigated the performance of various metaheuristic algorithms in Procedural Content Generation, specifically focusing on generating game map layouts. It challenges the common use of Genetic Algorithms by exploring alternative optimization approaches for PCG tasks. Three metaheuristic algorithms: Genetic algorithms, Particle Swarm Optimization, and Artificial Bee Colony were compared for their effectiveness and efficiency in generating game levels. Comprehensive experiments were conducted by using each algorithm for map creation. Metrics such as the convergence speed and content quality were used to evaluate the generated maps and identify the strengths and weaknesses of each algorithm in this context. The findings reveal that GA using tournament selection outperform other GAs implementations such as ABC and PSO for generating game maps. This leads to the conclusion that the choice of the optimal metaheuristic algorithm for PCG depends on the specific task and the selected evaluation methods. In addition, this study suggests that integrating diverse metaheuristic approaches can enrich the field of PCG and inspire game developers to utilize a broader range of techniques. By incorporating various algorithms, developers can potentially enhance content generation and create more immersive and engaging game experiences.

Time series data, which consists of observations collected sequentially over time, plays a pivotal role in various fields, including finance, economics, and environmental science. This paper delves into the mathematical foundations underpinning time series data, focusing on key concepts such as stationarity, autocorrelation, and time series models. By exploring how time series analysis is used to predict weather patterns and analyze long-term climate trends, this paper illustrates the power of mathematical modeling in uncovering insights about the world around us. Notably, the application of time series models in weather forecasting has revealed vital information about phenomena like El Niño and La Niña, which have profound impacts on global climate. This work underscores the significance of time series analysis in understanding and addressing environmental challenges.