In recent years, the proliferation of misinformation on social media platforms has become a significant concern. Initially designed for sharing information and fostering social connections, platforms like Twitter (now rebranded as X) have also unfortunately become conduits for spreading misinformation. To mitigate this, these platforms have implemented various mechanisms, including the recent suggestion to use crowd-sourced non-expert fact-checkers to enhance the scalability and efficiency of content vetting. An example of this is the introduction of Community Notes on Twitter. While previous research has extensively explored various aspects of Twitter tweets, such as information diffusion, sentiment analytics and opinion summarization, there has been a limited focus on the specific feature of Twitter Community Notes, despite its potential role in crowd-sourced fact-checking. Prior research on Twitter Community Notes has involved empirical analysis of the feature’s dataset and comparative studies that also include other methods like expert fact-checking. Distinguishing itself from prior works, our study covers a multi-faceted analysis of sources and audience perception within Community Notes. We find that contributors most often cite mainstream news and institutional sources with moderate or left-center bias and high factuality. Right-leaning or lower-factuality outlets appear more often in supportive notes, while highly factual and neutral sources are used to refute claims. Notes citing credible sources receive higher agreement, suggesting that Community Notes effectively rewards accuracy and reliability.
This study introduces novel approaches, namely the Jacobi Spectral Galerkin (JSG) and Iterated Jacobi Spectral Galerkin (IJSG) techniques, designed specifically to solve a system of linear Volterra integral equations (SLVIEs). These equations involve mixed-type kernels, incorporating both weakly singular (WS) and smooth kernels simultaneously. Jacobi polynomial-based Galerkin and Iterated Galerkin (IG) techniques have been used to tackle these integral equations. Initially, the existence and uniqueness of solutions for both the Galerkin and IG methods are established. Later, the convergence analysis is carried out for smooth and non smooth solution cases. The major attraction of this paper is its exploration of non-smooth solutions, alongside achieving superconvergent results, distinguishing it from other articles. Improved convergence rates are achieved by the IJSG method over the JSG method. The theoretical results are numerically validated.
The rapid growth of the Internet of Things (IoT) has transformed domains such as smart homes, healthcare, and industrial automation, while increasing concerns about security, privacy, and trust. Due to their decentralized structure and large-scale data generation, IoT systems are highly vulnerable to cyber threats, making transparency essential. This survey highlights Federated Learning (FL) and Blockchain as effective solutions. FL enables collaborative model training without sharing raw data, preserving privacy by exchanging only aggregated updates. Blockchain provides a distributed and immutable ledger that ensures transparent, verifiable, and tamper-resistant records. Together, they create a complementary framework where FL supports privacy-preserving intelligence and Blockchain ensures trust. The survey reviews their integration in IoT applications, examines existing research, and identifies key challenges such as scalability, computational overhead, and regulatory issues. It also outlines future research directions to support the development of secure, efficient, and trustworthy IoT systems. Furthermore, the work also surveys the system models, and experimental studies.
Alzheimer's Disease (AD) is a progressive neurodegenerative disorder characterized by memory loss and cognitive decline. Early and accurate detection of various stages of AD is crucial for effective treatment and care. Automatic detection of AD from MRIs has already been studied in the recent past. However, it remains a challenge due to the subtle and varied changes in the brain MRIs occurring at different stages of AD. This paper presents a unique DINOv2KAN, which strategically integrates the DINOv2 Vision Transformer (DINOv2) with Kolmogorov-Arnold Networks (KAN). The DINOv2 Vision Transformer extracts an initial feature representation from ADs' MRI scans. Subsequently, the KAN transforms these features into ADs' stage-specific robust feature representations used for identifying different stages of ADs. We employ a supervised contrastive learning approach to perform an end-to-end training of the proposed DINOv2KAN. Extensive experiments were conducted to evaluate the performance of DINOv2KAN, benchmarking it against other competing methods on three different widely used publicly available datasets: ADNI, OASIS and Kaggle. Our approach demonstrated superior performance, achieving an accuracy of 99.47% +/- 0.28% for AD vs. non-cognitively impaired classification and 99.11% +/- 0.25% for a four-way classification of AD on the ADNI dataset. Our model significantly outperforms state-of-the-art methods by at least 1.03% (p < 0.0001), 3.42% (p < 0.0001), and 2.58% (p < 0.0001) in terms of accuracy on ADNI, OASIS and Kaggle datasets, respectively. Furthermore, our model stands out best in the repeatability test. The source code is available at https://github.com/lahirisoham2004/DINOv2-KAN.
This letter proposes a Stackelberg game-theoretic framework for intelligent network selection in hybrid RF-VLC systems, addressing the demand for high-performance, cost-efficient wireless connectivity. Unlike conventional static allocation schemes, the framework enables dynamic decision-making where network providers, acting as leaders, optimize pricing, bandwidth, and transmission power, and users, as followers, select networks to maximize multi-dimensional utility functions. The approach jointly considers data rate, latency, energy efficiency, and reliability for users while balancing provider objectives of revenue and load distribution. Channel modeling uses the Lambertian model for VLC and log-distance path loss for RF, enabling accurate SNR and throughput estimation. Simulation results demonstrate notable gains in network efficiency, load balancing, and user satisfaction, highlighting the framework's scalability, interpretability, and computational efficiency over purely learning-based methods.
