We study the problem of resource reservation and allocation (RRA), aiming to enable network operators (NOs) to effectively develop the coexistence of enhanced mobile broadband (eMBB) and ultra-reliable low-latency (URLLC) services by acquiring resources-such as spectrum, base stations (BSs), and transmit power of the BSs-from the infrastructure provider (InP). To do this, we quantitatively analyze the impact of spatial-temporal aspects of the network, including the spatial distributions of users and BSs, wireless channel conditions, and user traffic statistics, on resources required to support users' requirements using the meta distribution. Since spatial-temporal aspects operate on different time and space scales, we investigate a multi-timescale RRA (MT-RRA) scheme that facilitates collaboration between the InP and NOs to acquire resources through long-timescale RR (LT-RR) over the spatial distributions of users and user traffic statistics and medium-timescale RR (MT-RR) over wireless channel fadings. Subsequently, NOs allocate these resources to their eMBB and URLLC users in shorter timescales. To address these problems, an iterative decomposition method is proposed. The MT-RRA scheme can take advantage of the spatial-temporal aspects, flexibility of the MT-RR, and costeffectiveness of the LT-RR. Our simulation results demonstrate that the MT-RRA scheme can obtain 10% and 23% improvements in the average throughput of eMBB and URLLC users, respectively, regarding the link reliability of the eMBB and URLLC users in comparison with single-timescale RRA schemes.

In emergency scenarios with incapacitated (compromised/absent) fixed communication infrastructure, ensuring reliable on-demand and dynamic network coverage becomes critical for rapid and effective disaster management. This paper proposes a novel framework for deploying network-in-a-box (NIB) solution using a hybrid model that integrates groundbased NIBs for accessible areas with airborne NIBs for areas with restricted ground access. The proposed approach leverages the portability, compactness, plug-and-play characteristics of NIBs, flexibility/mobility of unmanned aerial vehicle (UAV) based NIBs, and reliability/stability of ground-based NIBs to establish a temporary yet scalable, adaptive, and dynamic communication network. We adopt priority-based NOMA (PB-NOMA) for multiuser multi-antenna downlink (DL) and uplink (UL) communications between NIBs and single-antenna ground users. We further employ simultaneous wireless information and power transfer (SWIPT) for priority users (PUs) to replenish their equipment batteries for continued operation while transferring wireless information to general users (GUs). Our proposed approach addresses the key challenges such as terrain-aware rapid deployment, dynamic resource allocation, complete/partial power outages, and seamless connectivity with the aim to maximize generalized global energy efficiency (GGEE). Our analysis, optimization framework, and simulation results demonstrate the effectiveness of the adopted system to provide robust ondemand critical communications while dynamically adapting to user traffic demands, environmental constraints, and emergency scenarios.

Many people use online platforms such as Instagram and Facebook on a daily basis. These applications often feature comment sections where users can express their opinions. Unfortunately, it is common for this feature to be used inappropriately, with people posting offensive comments towards individuals or groups. Defining what represents offensive content is difficult, as it varies by context, making AI removal challenging. In order to offer additional awareness on hateful and abusive conduct in social media, this work focuses on the SemEval-2023 task of binary sexism detection. Our data analysis indicates that adding a new label would be beneficial for enhancing the dataset's alignment with real-world applications. Towards this end, this work presents a simple baseline model using cosine similarity to assess comment alignment with different platform guidelines to highlight how these comments differ from the community guidelines.

With the popularization of Bluetooth technology, its security issues become especially critical. As a typical security threat, Bluetooth man-in-the-middle attack can cause serious damage to the integrity and privacy of data transmission. To address this challenge, this paper proposes a real-time Bluetooth signal anomaly detection model based on graph neural network (GNN). The model focuses on detecting and identifying potential man-in-the-middle attacks from complex Bluetooth signals. We first construct a Bluetooth signal dataset containing rich scenarios that model diverse communication environments and potential attack patterns. Using a graph representation learning approach, we convert the signals into graph structures, enabling the use of graph topology to analyze signal properties. Then, we design a novel graph convolution operation that captures spatio-temporal features in the signals and efficiently extracts patterns associated with attack behaviors. In the experimental section, we demonstrate the significant advantages of the proposed model in terms of detection accuracy, robustness and real-time performance by comparing it with existing methods. The model not only exhibits high accuracy in anomaly detection on the dataset, but also meets the demand of real-time monitoring in terms of processing speed. In addition, we also evaluate the generalization ability of the model and demonstrate its stability and reliability on different environments and devices.

This paper serves as a comprehensive guide for practitioners and scholars aiming to understand the channel coding and decoding schemes integral to the 5G NR standard, with a particular focus on LDPC and polar codes. We start by explaining the design procedures that underlie these channel codes, offering fundamental information from perspectives of both encoding and decoding. In order to determine the present status of research in this area, we also provide a thorough literature review. Notably, we add comprehensive, standard-specific information to these foundational evaluations that is frequently difficult to extract from technical specification documents.

As wireless radio frequency (RF)-based localization techniques continue to attract interest, a plethora of approaches including received signal strength (RSSI) trilateration and multi-lateration, phase, time-of-arrival, and machine learning models have been explored for indoor localization. However, there has been no comprehensive experimental investigations that compared the accuracy of these methods in a practical IoT wireless sensor network (WSN). Herein, we present a holistic evaluation of localization techniques in an indoor smart home environment, based on off-the-shelf 868/915 MHz transceivers. First, the hardware limitations such as the antenna and RSSI radiation patterns and the effects of multi-path reflections are experimentally investigated, identifying the optimal node placement. A practical RSSI recording and forwarding scheme is proposed and implemented using microcontroller units (MCUs), showing a frugal approach for Joint Sensing and Communication (JSAC), with under 420 ms cycle time. Using this testbed, we compare multi-lateration approaches for 3 and 4 receivers, in both line-of-sight (LOS) and non-line-of-sight (NLOS) links, achieving between 46% and 89% room prediction accuracy, with a minimum mean error of 1.49 m. A machine learning (ML)-based approach, using multinomial logistic regression, is then reported with a peak room classification accuracy of 97-100%, for 25-30 RSSI points. A comparison with state-of-the-art implementations is presented showing a high room localization accuracy at a low hardware complexity, demonstrating the feasibility of RSSI-only localization in resourceconstrained IoT networks.

Mixture of Experts (MoE) models have emerged as a powerful framework in machine learning, offering a scalable and efficient approach to solving complex tasks by leveraging a mixture of specialized models, or "experts," each of which is responsible for different aspects of the input data. By activating only a small subset of experts at a time, MoE models achieve high performance while maintaining computational efficiency, making them particularly well-suited for large-scale applications. This survey provides an overview of the MoE model, discussing its foundational principles, recent advancements, and current challenges. We examine the role of the gating mechanism, the trade-off between model complexity and computational efficiency, and the challenges of training stability, expert redundancy, and scalability. Additionally, we highlight recent innovations in sparse activation, hybrid architectures, and integration with metalearning and self-supervised learning. Despite the progress made, challenges related to interpretability, fairness, and deployment in real-world systems remain. Finally, we explore future directions for MoE research, including smarter expert selection, multimodal learning, and edge deployment. This survey aims to provide a comprehensive understanding of MoE models and their potential to drive future advancements in artificial intelligence.

Integrated sensing and communication (ISAC) has emerged as a promising technology that enables the simultaneous operation of sensing and communication functions. By sharing hardware infrastructure, spectrum resources, and signal waveforms, ISAC can reduce hardware costs and enhance spectral efficiency. However, existing studies have ignored the interference caused by echo signals from untargeted sensing objects (SOs) in large-scale ISAC networks. Therefore, this paper investigates the ISAC coverage probability, defined as the weighted average of the probabilities that communication and sensing signal-tointerference ratios exceed their corresponding thresholds, in large-scale ISAC networks with multiple SOs. Firstly, the sensing channel gains are approximated by Gamma random variables using a moment matching method. Secondly, the analytical expressions of communication and sensing coverage probabilities are derived using stochastic geometry and validated by Monte Carlo simulations. Equipped with these results, we numerically analyze the effects of the number of antennas at base stations (BSs), the BS density, and the numbers of communication users (CUs) and SOs on the ISAC coverage probability. The numerical results show that the echo-signal interference from untargeted SOs significantly impacts the ISAC coverage probability, which decreases the ISAC coverage probability by up to 19% compared to scenarios without this interference. This indicates that the echo-signal interference cannot be ignored in ISAC networks, and its negative impact can be significantly reduced by increasing receiving antenna numbers at BSs.

Traditional spatiotemporal data analysis often relies on predictive models, which overlook the causal relationships driving observed phenomena. This limitation makes it hard to identify true drivers and develop effective interventions. To address this, we explore causal machine learning techniques for spatiotemporal data, aiming to provide robust and interpretable insights. Our review of the literature found that less than 1% of studies, based on rigorous criteria, explicitly combine causal machine learning with spatiotemporal data analysis. Most of these studies focus on improving causal effect discovery and estimation (32 papers), enhancing prediction accuracy (19 papers), and addressing pattern recognition limitations (10 papers). Surprisingly, only one paper discusses using causal machine learning to improve algorithm interpretability for spatiotemporal data, highlighting a major gap. We examine the unique aspects of spatiotemporal data, such as spatial autocorrelation and temporal dependencies, which present both opportunities and challenges for causal inference. Specific methods like spatiotemporal Granger causality and structural equation modeling with spatial lags show potential in capturing complex interdependencies and providing interpretable results. Building on these insights, we propose novel conceptual approaches to model underlying causal structures and suggest promising research directions, such as developing interpretable models and advancing real-time causal inference algorithms in dynamic environments. We also discuss computational challenges like scalability, efficiency, and the trade-offs between model complexity and interpretability that need to be addressed for practical applications. Ethical considerations, including mitigating biases in causal discovery and their societal implications, are also covered. This research contributes to a comprehensive framework for leveraging causal machine learning in spatiotemporal data analysis, with applications in fields like climate science, economics, epidemiology, and urban planning.

With the rapid development of the retail industry, enhancing customer experience and operational efficiency has become increasingly critical, where technological integration is key. This study introduces an innovative framework for human behavior recognition that combines GNN and RFID technology. By embedding RFID signals into the graph structure, we effectively capture the spatial dependencies hidden in the data. Furthermore, the Spline Convolution technique is utilized to address the spatial dependencies of the signals, achieving accurate and robust human behavior recognition. Facing challenges such as dynamic changes in data dimensions in the retail environment, over-smoothing issues in graph neural networks, and the effective fusion of multi-dimensional features, we adopted a graphbased modeling approach. We constructed an adjacency matrix with small-world characteristics using the TopK mechanism and Pearson correlation coefficients, and introduced Inception structures and residual connections to increase network width, thereby mitigating over-smoothing phenomena. The introduction of BiLSTM readout methods further enhanced the model's ability to process time series information. Experimental results demonstrate the framework's excellent performance in human behavior recognition tasks, with high accuracy and strong robustness, proving not only theoretically feasible but also highly effective in practical applications. Through qualitative analysis, we have improved the interpretability of the framework, providing retailers with a powerful tool for gaining in-depth insights into customer behavior, which helps to optimize customer experience and enhance operational efficiency.

Differential semantic communication represents an intelligent model for meaning-based interactions between a transmitter and a receiver through a closed-loop feedback system. However, classical information theory, underpinned by Shannon's research, lacks the generality to analyse meaning-driven communications in complex feedbacked systems. To address this limitation, we propose a novel information theory for modern communication networks. Our theory introduces the concepts of transmitter and receiver capacities, which complement Shannon's channel capacity. We demonstrate how information redundancy can be strategically identified to align with communication goals. To validate the theory, we quantify key metrics to compare various image coding schemes. Our proposed work provides a theoretical foundation for advancing the design and optimisation of next-generation semantic communication systems.

Transformers have emerged as the foundation of modern deep learning, achieving state-of-the-art performance across diverse domains such as natural language processing, computer vision, and multimodal tasks. However, their quadratic computational complexity and high memory requirements pose significant challenges, especially when scaling to long sequences or deploying on resource-constrained hardware. To address these limitations, the research community has proposed a variety of techniques to make Transformers more efficient. This survey provides a comprehensive overview of these advancements, categorizing them into five main strategies: sparsity-based methods, low-rank approximations, structured attention patterns, memory-efficient mechanisms, and hybrid approaches. For each category, we delve into the underlying principles, highlight representative methods, and analyze their strengths and limitations. We also discuss the trade-offs involved, including computational efficiency, memory savings, and model accuracy. Furthermore, we identify emerging trends and open challenges, such as unified frameworks, hardware-aware optimizations, and sustainable AI practices. By synthesizing the state of the art in efficient Transformers, this survey aims to guide future research and facilitate the adoption of these techniques in both academic and industrial settings.

In this paper, we introduce the Chaotic Dynamic Cat Swarm Optimization (CHDCSO) algorithm, an enhanced variant of the traditional Cat Swarm Optimization (CSO), to address the Traveling Salesman Problem (TSP). By integrating chaotic dynamics using logistic chaotic maps, the proposed algorithm aims to improve the balance between exploration and exploitation phases, thereby enhancing the global search capability and robustness of the optimization process. Detailed experimental analysis is conducted on various TSP datasets, comparing the performance of CHDCSO against established optimization techniques, including Particle Swarm Optimization (PSO), Differential Evolution (DDE), Transit Search, Random Optimization Algorithm (ROA), and Random Search Algorithm (RSA). The results demonstrate that the CHDCSO algorithm consistently outperforms these traditional methods in terms of convergence speed and solution quality. This study underscores the potential of chaotic dynamics in optimizing complex problems and highlights the CHDCSO algorithm’s efficacy in achieving superior optimization outcomes.

This paper introduces a novel framework that leverages ontology engineering to enhance big data science, addressing challenges such as data heterogeneity, integration, and scalability. The framework employs semantic technologies to unify diverse datasets, enabling advanced knowledge representation, reasoning, and predictive modelling. By integrating ontological methods with big data analytics, the framework demonstrates improved data integration and analysis across multiple domains. Case studies in big data management and healthcare systems illustrate the practical applications and benefits of the framework, including enhanced data quality, semantic alignment, and process efficiency. This work advances the field of big data science by providing a scalable and interoperable solution for managing complex datasets, paving the way for future research and applications in data-driven domains.

Doppler Weather Radars (DWRs) play a crucial role in atmospheric monitoring by providing high-resolution precipitation and wind data. However, electromagnetic interference (EMI) can compromise signal quality, reducing accuracy. Traditional band-pass filters used in DWRs often have a wide frequency range, allowing unwanted signals to pass through and degrading performance. The increasing demand for compact, high-performance filters in telecommunications and radar systems necessitates innovative design approaches.

An improved understanding of pest population dynamics is critical for more informed and effective management of agricultural pests. A Digital Twin (DT) that spatially visualises pest population dynamic modelling and monitoring, has great potential for facilitating effective and collaborative pest management. We have developed a proofof-concept DT of agricultural pest monitoring that provides dynamic and real-time estimates of pest population throughout an agricultural landscape, while accounting for weather, landuse and pest control methods being applied. This proof-ofconcept DT demonstrates both the viability and suitability of the DT framework for agricultural pest monitoring and management; it also showcases the potential of DTs as valuable decision-support tools in the agricultural landscape.

Enterprises face increasing challenges in managing authentication across thousands of applications while ensuring security and compliance. This article explores a scalable authentication framework that integrates modern identity providers with internal systems, enabling seamless transitions from legacy authentication. By leveraging centralized access control and adaptive authentication, organizations can enhance security while maintaining operational efficiency. This article shares key lessons from implementing this architecture in largescale enterprises and provides practical strategies for security teams to modernize authentication without disrupting workflows.

In this paper, we provide informed arguments for using columnar in-memory database technology, in particular the Hyrise database, for the high-performant implementation of the highly combinatorial data mining tool GrandReport. In service of that we provide a targeted review of columnar databases, a targeted review of columnar in-memory databases, and a comparison of nine established columnar in-memory databases. On the basis of this, we discuss advantages of using the Hyrise database as the future implementation platform for the GrandReport tool.