The k-Nearest-Neighbors (kNN) search in the high-dimensional space is a fundamental problem in many applications. However, retrieving the nearest neighbors from a large-scale and high-dimensional dataset is computationally challenging, i.e, existing approaches either suffer from high construction cost or unsatisfactory search performance. In this paper, we propose a fast and light-weight learned index FLEX for kNN search in high-dimensional space. First, we develop a multi-module DNN framework that needs much less training examples than employing classical DNN models directly to reach the same accuracy level. Second, we propose a linear-time data layout algorithm which aims to maximize the accuracy under a search cost constraint. A bound of the expected approximation ratio of the proposed algorithm is also proved. Experimental evaluation shows that our index significantly outperforms the state-of-the-art indexes in terms of both time efficiency and space efficiency while maintaining competitive or better accuracy performance.

As a prospective key technology for the next-generation wireless communications, reconfigurable intelligent surfaces (RISs) have gained tremendous research interest in both the academia and industry in recent years. Only limited knowledge, however, has been obtained about the channel eigenvalue characteristics and spatial degrees of freedom (DoF) of systems containing RISs, especially when mutual coupling (MC) is present between the array elements. In this paper, we focus on the small-scale spatial correlation and eigenvalue properties excluding and including MC effects, for RISs with a quasicontinuous aperture (i.e., holographic RISs). Specifically, asymptotic behaviors of far-field and nearfield eigenvalues of the spatial correlation matrix of holographic RISs without MC are first investigated, where the counter-intuitive observation of a lower DoF with more elements is explained by leveraging the power spectrum of the spatial correlation function. Second, a novel metric is proposed to quantify the inter-element correlation or coupling strength in RISs and ordinary antenna arrays. Furthermore, in-depth analysis is performed regarding the MC effects on array gain, effective spatial correlation, and eigenvalue architectures for a variety of element intervals when a holographic RIS works in the radiation and reception mode, respectively. The analysis and numerical results demonstrate that a considerable amount of the eigenvalues of the spatial correlation matrix correspond to evanescent waves that are promising for near-field communication and sensing. More importantly, holographic RISs can potentially reach an array gain conspicuously larger than conventional arrays by exploiting MC, and MC has discrepant impacts on the effective spatial correlation and eigenvalue structures at the transmitter and receiver.

The SyncFusion Amplifier is introduced as an innovative electronic component designed to redefine security in IoT devices. Principles inspired by quantum entanglement are leveraged to establish synchronized correlations between input signals, resulting in amplified and coherent outputs. The behavior, validation, and potential applications of the amplifier are demonstrated using QSPICE simulation and a hierarchical entry block with C++ code. Through detailed analysis, the enhancement of correlated behaviors by the SyncFusion Amplifier is showcased, with applications in cryptography and communication. By generating more robust cryptographic keys and secure communication channels, security vulnerabilities in IoT environments are addressed. The research not only unveils a groundbreaking electronic component but also proposes a paradigm shift in IoT security. As the IoT landscape continues to expand, a timely solution is presented by the SyncFusion Amplifier to fortify data protection, communication integrity, and overall device security.

If tokenized assets are to be a reality in the future decentralized Web3 then transaction keys need to be distributed and under the control of the asset-owners. This requires a careful design of wallet systems based on trusted hardware. A core feature needed for wallet systems is the attestation of the state of the transaction keys in the wallet without disclosure of the keys. This feature is relevant for relying parties such as insurance providers who need to perform risk assessment based on the security quality of the environment inside the wallet system that is protecting the transaction keys. In the longer term, all key-bearing devices that participate in a decentralized tokenized asset network will need to be hardened using trusted hardware, with attestation capabilities for detecting and countering cyberattacks.

Train motion model calibration is crucial to ensure the accuracy and performance of any railway algorithm that requires reproducing and predicting train dynamics. The calibration must be robust against measurement outliers and missing data. However, such a robust train motion model calibrator is still missing in the literature. In this paper, we propose a robust parameter estimator based on M-estimation for calibrating the train motion model using historical location, speed and engine torque measurements. M-estimation utilizes an objective function that reduces the influence of outliers on the estimation, making the estimates more robust. The estimation problem is solved iteratively as a weighted least squares regression. We tested 10 error measures, finding that Fair's is the one performing best for estimating the running resistance. The maximum tractive and brake effort and the braking style of drivers were also quantified by filtering effort measurements and deceleration rates with a sliding median filter. We show the performance and robustness of the proposed method in a real case study, using data from trains running in the Dutch railway network. The statistics obtained show that M-estimation outperforms Ordinary Least Squares and that in manually-driven trains the accuracy of any train motion model relies heavily on the actual characteristics of manual driving.

This paper presents the development and implementation of a low-cost, portable scale utilizing a smartphone and image  processing technology. The main objective of this research is to explore the potential of using a compliant mechanism, markers, and  image processing techniques to accurately determine the weight of an object. We designed a 3D printed platform fitted with a  webcam to capture images of white markers on a black surface. The Image Processing Toolbox on MATLAB was used to detect  two parameters, the number of white pixels and the centroid positions of the white pixels with respect to the centre of the image in  each marker. Calibration is carried out to convert these two parameters into weight, hence accurately estimating the weight of an  object placed on the platform.

An off-line Nepali numeral handwritten character recognition, based on the neural network, is described in this paper. Nepali numeral specific features are extracted from character images. Connection of feature vector to the hidden neurons and hidden neurons to the output neurons is chosen to give optimum result. Very few number of samples are used to train the neural network. Recognition accuracy of 81.14 per cent is obtained.

The demand for more accurate and visually pleasing images is increasing with the number of digital photos taken daily. However, the images captured by modern cameras are degraded by noise, which leads to deteriorated visual image quality. The de-noising approaches existing in the literature cannot reduce noise without losing image features (edges, corners, and other sharp structures). The developed de-noising method can detect and de-noise the noisy pixels without losing image features at higher and lower noise levels. The present approach can filter the noise at or above 90% noise density and preserve image details to enhance the image processing task.

Probabilistic computing with binary stochastic neurons (BSN) implemented with low- or zero-energy barrier nanoscale ferromagnets (LBMs) possessing in-plane magnetic anisotropy has emerged as an efficient paradigm for solving computationally hard problems. The fluctuating magnetization of an LBM at room temperature encodes a p-bit which is the building block of a BSN. Its only drawback is that the dynamics of common (transition metal) ferromagnets are relatively slow and hence the number of uncorrelated p-bits that can be generated per second – the so-called “flips per second” (fps) – is insufficient, leading to slow computational speed in autonomous co-processing with p-computers. Here, we show that a simple way to increase fps is to replace commonly used ferromagnets (e.g. Co, Fe, Ni), which have large saturation magnetization Ms, with a dilute magnetic semiconductor like GaMnAs with much smaller saturation magnetization. The smaller Ms reduces the energy barrier within the LBM and increases the fps significantly. It also offers other benefits such as increased packing density for increased parallelization and reduced device to device variation. This provides a way to realize the hardware acceleration and energy efficiency promise of p-computers.

This paper presents a monolithic, inexpensive, label free, robust pH sensor that can be fabricated in an unmodified CMOS process. The sensor uses a differential p-ISFET architecture which substantially reduces non-ideal common mode effects which corrupt readout in ISFET based pH sensors and occupy only a small footprint. Additionally, the design offers the capability to correct for drift through control of the n-well voltage, which has not been demonstrated previously. We present results obtained from testing the sensor that is fabricated in the AMS 0.35μm process, including a demonstration of the drift correction. We demonstrate a differential sensitivity of 15 mV/pH over the entire pH range, with the ability to remove the already small drift of 0.1 mV/min using the bulk-source voltage adjustment.

Existing multimodal sentiment analysis models focus more on fusing highly correlated image-text pairs, and thus achieves unsatisfactory performance on multimodal social media data which usually manifests weak correlations between different modalities. To address this issue, we first build a large multimodal social media sentiment analysis dataset RU-Senti which contains more than 100,000 image-text pairs with sentiment labels. Then, we proposed a topic-oriented model (TOM) which assumes that text is usually related to a certain portion of the image contents and the image-text pairs of the same topic often have similar sentiment tendencies. TOM learns the topic information from textual content and designs a topic-oriented feature alignment module to extract textual semantics correlated information from images, thus achieving the alignment between two modalities. Then, TOM utilizes a transformer encoder initialized with the parameters from a pre-trained vision-language model to fuse the multimodal features for sentiment prediction. According to the experiments over the public MVSA-Multiple dataset and our RU-Senti dataset, RU-Senti is of high suitability for studying weakly correlated multimodal sentiment analysis, and the proposed TOM model also largely outperforms the SOTA mulitimodal sentiment analysis methods and pre-trained vision-language models. The RU-Senti dataset and the code of TOM are available at https://github.com/PhenoixYANG/TOM.

Drones are used for public safety missions because of their communication capabilities, unmanned mission, flexible deployment, and low cost. Recently, drone-assisted emergency communication systems in disasters have been developed where instead of a single large drone, flying ad hoc networks (FANETs) are proposed through clustering. Although cluster size has an impact on the proposed system’s performance, no method is provided to effectively regulate cluster size. In this paper, optimum cluster size is obtained through two distinct meta-heuristic optimization algorithms - the Cuckoo Search Algorithm (CUCO) and the Particle Swarm Algorithm (PSO). Flowcharts and algorithms of CUCO and PSO are provided. A presentation of an analytical investigation based on the Markov chain model is provided. To further validate the analytical study, simulation results are presented. Simulation shows the improvement in terms of throughput and packet dropping rate (PDR).

As Metaverse emerges as the next-generation Internet paradigm, the ability to efficiently generate content is paramount. AIGenerated Content (AIGC) emerges as a key solution, yet the resourceintensive nature of large Generative AI (GAI) models presents challenges. To address this issue, we introduce an AIGC-as-a-Service (AaaS) architecture, which deploys AIGC models in wireless edge networks to ensure broad AIGC services accessibility for Metaverse users. Nonetheless, an important aspect of providing personalized user experiences requires carefully selecting AIGC Service Providers (ASPs) capable of effectively executing user tasks, which is complicated by environmental uncertainty and variability. Addressing this gap in current research, we introduce the AI-Generated Optimal Decision (AGOD) algorithm, a diffusion model-based approach for generating the optimal ASP selection decisions. Integrating AGOD with Deep Reinforcement Learning (DRL), we develop the Deep Diffusion Soft Actor-Critic (D2SAC) algorithm, enhancing the efficiency and effectiveness of ASP selection. Our comprehensive experiments demonstrate that D2SAC outperforms seven leading DRL algorithms. Furthermore, the proposed AGOD algorithm has the potential for extension to various optimization problems in wireless networks, positioning it as a promising approach for future research on AIGC-driven services. The implementation of our proposed method is available at: https://github.com/Lizonghang/AGOD.

Email spam, characterized by unsolicited and often harmful messages, remains a pervasive issue across digital communication platforms. This study explores advanced Machine Learning (ML) and Deep Learning (DL) methodologies to enhance email spam detection, aiming to protect users from phishing, malware, and other malicious threats. A comprehensive literature review is conducted, detailing the processes involved in developing spam detection models, including preprocessing, feature extraction, and model implementation. This study focuses on three ML algorithms-Random Forest (RF), Naive Bayes (NB), and Support Vector Machine (SVM)-and three DL models-Long Short-Term Memory (LSTM), Bidirectional Long Short-Term Memory (Bi-LSTM), and Bidirectional Encoder Representations from Transformers (BERT). A comparative analysis of these models is presented, highlighting their accuracy and performance. Notably, DL models, particularly BERT, achieved the highest accuracy (up to 99.14%), while ML models like SVM also demonstrated strong performance (up to 99.00%). The findings indicate that DL models outperform traditional ML algorithms in accuracy, although ML models offer advantages in interpretability and computational efficiency. This study concludes with insights into the strengths and weaknesses of both approaches and suggests future research directions to enhance the robustness and effectiveness of spam detection systems.

In recent years, the likelihood of wildfire occurrence has increased in many North American communities as changes in climate have led to longer, more deadly fire seasons.  Many Americans, especially those living in Western states, have reported frequent drought and wildfire conditions, leading to an increased need for a modeling program to assess wildfire risk at a low computational cost. The research objective of this paper was to develop a machine learning model capable of producing real-time wildfire risk assessments using five geospatial datasets: Land Fire Mean Return, Annual Precipitation, Sentinel-2 Imagery, Land Cover, and Moisture Deficit & Surplus. To create the model, three separate machine learning architectures were implemented (U-Net, DeepLabV3, and the Pyramid Scene Parsing Network) and then applied to the study area of San Bernardino County, CA for the year 2020. In addition, this study demonstrated a proof of concept for further inquiry into combining artificial intelligence and geospatial datasets to create useful insights. Â

This study showcases Convolutional Neural Networks’ (CNNs) potential in detecting breast cancer from ultrasound images. Impressive accuracy highlights their value in precise and efficient diagnosis, paving the way for further research. The CNN’s remarkable accuracy, even with a small dataset of 1,500 images, suggests enhanced performance with larger and diverse datasets. CNNs for breast cancer detection via ultrasound offer a promising, potent approach to revolutionize medical imaging and diagnosis, boosting accuracy and reliability.

This paper investigates the concept of frequency in arbitrary multi-phase systems based on geometrical principles. The proposed approach relies on state-of-the-art mathematical techniques such as differential geometry and geometric algebra in $\bm{n}$ dimensions. By analyzing the generalized Frénet-Serret frame, we derive how the Darboux bivector can accurately express the rotation of this frame as a rigid body in space. It is shown how the concept of frequency in power grids can be intimately linked to spatial rotations. New insights are presented based on the comparison with other recently published works. It is also concluded that the application to single-phase systems cannot always be accommodated by spatial curves. Several examples are used to illustrate the findings of this paper.

A novel circuit-computational electromagnetic (CEM) hybrid simulation methodology, which results in vast improvements in the accuracy of high-frequency photovoltaic (PV) modelling between 1 Hz and 50 MHz, is proposed. At low frequencies, the impedance of a PV module is dominated by the behaviour of the PV cells, while at high frequencies it is dominated by the layout of the inter-cell connections. The proposed methodology considers both the low and high-frequency behaviour of a PV module in a single simulation. Firstly, the impedances of two PV modules are carefully measured using two vector network analyser configurations. Secondly, literature-based models of PV modules are adapted to the proposed methodology. Thirdly, improved models of each PV module are introduced in four ascending levels of complexity. These improved models comprise additional novel aspects, including distributed cell layouts and implicit CEM-calculated self-inductances. Fourthly, a feature-selective method is used to compare the measured results with the literature-based and improved models. This analysis yielded comparable error reductions of up to a factor of 400 – highlighting the value of the improved methodology and models. Penultimately, the required level of computational resources is introduced, and the trade-off between simulation time and accuracy is discussed. Finally, a review discussion is presented.