Attribute reduction is a core technique in the rough set domain and an important step in data preprocessing. Researchers have proposed numerous innovative methods to enhance the capability of attribute reduction, such as the emergence of multi-granularity rough set models, which can effectively process distributed and multi-granularity data. However, these innovative methods still have many shortcomings, such as how to deal with complex constraints and how to perform multi-angle effectiveness evaluations. Based on the multi-granularity model, this study proposes a new method of attribute reduction, namely using multi-granularity neighborhood information gain ratio as the measurement criterion. This method combines both supervised and unsupervised perspectives, and by integrating multi-granularity technology with neighborhood rough set theory, constructs a model that can adapt to multi-level data features. This model can select the optimal granularity level or attribute set according to the requirements. Finally, the proposed method is experimented with on 15 UCI datasets against six other attribute reduction algorithms. The experimental outcomes, upon analysis and comparative review, substantiate the reliability and consistency of the attribute reduction approach introduced in this study.

The work aims to develop an electromagnetic shielding measurement methodology and design an electromagnetic shielding effectiveness test setup for thin sheet material testing. In brief, multiple shielding effectiveness standards are described and analyzed in the thin sheet material testing framework. Two types of the developed shielding effectiveness test probe in high and low-frequency ranges are validated comprehensively. Extensive descriptions and data of developed testing methodology are also reported.

Public safety and intelligent surveillance systems critically depend on anomaly behavior detection for effective monitoring. In real-world pedestrian detection scenarios, prevalent challenges such as missed detections, complex background interference, and small target sizes hinder accurate anomaly identification. To address these issues, this study introduces YOLO-ABD, a lightweight method for anomaly behavior detection that integrates small target detection and channel shuffling. This method employs YOLOv8n as the baseline model, incorporating a small target detection mechanism in the Head part and utilizing GSConv convolution in the Backbone to enhance perceptual capability. Additionally, the SimAM attention mechanism is integrated to mitigate complex background interference, thereby improving target detection performance. Evaluation on the IITB-Corridor dataset demonstrated mAP50 and mAP50-95 scores of 89.3% and 60.6%, respectively. Generalization testing on the street-view-gdogo dataset further highlighted the superiority of YOLO-ABD over advanced detection algorithms, underscoring its effectiveness and generalizability. With a relatively small parameter count, YOLO-ABD presents an excellent lightweight solution for pedestrian anomaly behavior detection.

In recent decades, with the ever-growing scale of video data, near-duplicate videos continue to emerge. Data quality issue caused by near-duplicate videos is becoming more and more prominent, which has affected the application of normal videos. Although current studies on near-duplicate video detection can be helpful to uncover data quality issues for videos, they still lack a process of automatic merging for the video data represented by high-dimensional features, which are difficult to automatically clean the near-duplicate videos to improve data quality for video datasets. At present, there are few studies on near-duplicate video data cleaning. The existing studies have the sensitive problems of video data orderliness and clustering initial center under a condition that prior distribution is unknown, which seriously affect the accuracy of near-duplicate video data cleaning. To address the above issues, an automatic near-duplicate video data cleaning method based on a consistent feature hash ring is proposed in this paper. First, a residual network with convolutional block attention modules, a long short-term memory deep network, and an attention model are integrated to construct an RCLA deep network with the multi-headed attention mechanism to extract spatiotemporal features of video data. Then, a consistent feature hash ring is constructed, which can effectively alleviate the sensitivity of video data orderliness while providing a condition of near-duplicate video merging. To reduce the sensitivity of the initial cluster centers to results of near-duplicate video cleansing, an optimized feature distance-means clustering algorithm is constructed by utilizing a mountain peak function on a consistent feature hash ring, which can implement automatic cleaning of near-duplicate video data. Finally, experiments are conducted based on a commonly used dataset named CC_WEB_VIDEO and a coal mining video dataset. Compared with some existing works, simulation results demonstrate that the performance of the proposed method.

Computer vision is a rapidly developing field that focuses on highly sophisticated picture analysis, manipulation, and comprehension. Its objective is to analyze what is happening in front of a camera and utilize that understanding to control a computer or robotic system or to present users with fresh visuals that are more enlightening or appealing than the original camera images. Computer vision technologies make it feasible for new user interfaces, augmented reality gaming, biometrics, automobile, photography, movie creation, Web search, and many more applications. This essay seeks to explain how computer vision can be utilized to play blackjack successfully.

In the intersection of the Internet of Things (IoT) and Machine Learning (ML), the choice between high-level and low-level programming libraries presents a significant dilemma for developers, impacting not only the efficiency and effectiveness of ML models but also their environmental footprint. We have proposed a comprehensive framework to aid in this decision-making process, underpinned by a detailed comparative analysis of both types of libraries on one of the key IoT ML task; image classification. We have introduced a novel algorithm designed to calculate the green footprint of ML model training, factoring in execution time, memory utilization, power consumption, and CPU temperature, addressing the urgent need for sustainable ML practices. Through an empirical evaluation of popular libraries such as PyTorch for high-level and Libtorch for low-level development, we have assessed their performance, development efficiency, and hardware compatibility. The culmination of our research is a decision support system that synthesizes the experimental findings to guide developers toward choices that harmonize model performance with environmental sustainability.

Day-ahead forecasting of electricity prices is important in deregulated electricity markets for all the stakeholders: energy wholesalers, traders, retailers, and consumers. Electricity price forecasting is an inherently difficult problem due to its special characteristic of dynamicity and non-stationarity. In this paper, we present a robust price forecasting mechanism that shows resilience towards aggregate demand response effect and provides highly accurate forecasted electricity prices to the stakeholders in a dynamic environment. We employ an ensemble prediction model in which a group of different algorithms participate in predicting the price for each hour of a day. We propose two different strategies, namely, Fixed Weight Method (FWM) and Varying Weight Method (VWM), for selecting each hour's expert algorithm from the set of participating algorithms. In addition, we utilize a carefully engineered set of features selected from a pool of features derived from information such as past electricity price data, weather data, and calendar data. The proposed ensemble model offers better results than both the Pattern Sequence-based Forecasting (PSF) method and our own previous work using Artificial Neural Networks (ANN) alone do on the datasets for New York, Australian, and Spanish electricity markets.

In the study of algebraic structures related to logical systems, Ideals and filters have different meanings and they are algebraic notions related to logical provable formulas. Unlike the classical Boolean lattice theory, ideals and filters are not dual notions in residuated lattices. An interesting subclass of residuated lattices is the class of triangle algebras, which is an equational representation of interval-valued residuated lattices that provides an algebraic framework for using closed intervals as truth values in fuzzy logic. The main aim of this article is to introduce and study the concept of ideals in triangle algebras and investigate the connection between ideals and filters. We first point out that the construction procedure for the filter generated by a subset of a triangle algebra established by Zahiri et al. is incorrect, and we proceed to give an alternative characterization.

Quantum computing (QC) has opened the door to advancements in machine learning (ML) tasks that are currently implemented in the classical domain. Convolutional neural networks (CNNs) are classical ML architectures that exploit data locality and possess a simpler structure than fully-connected multi-layer perceptrons (MLPs) without compromising the accuracy of classification. However, the concept of preserving data locality is usually overlooked in the existing quantum counterparts of CNNs, particularly for extracting multi-features in multidimensional data. In this paper, we present a multidimensional quantum convolutional classifier (MQCC) that performs multidimensional and multi-feature quantum convolution with average and Euclidean pooling and thus adapting the CNN structure to a variational quantum algorithm (VQA). Average pooling is based on the quantum Haar transform (QHT) and Euclidean pooling is based on partial quantum measurement. The experimental work was conducted using multidimensional data to validate the correctness and demonstrate the scalability of the proposed method utilizing both noisy and noise-free quantum simulations. We evaluated the MQCC model with reference to reported work on state-of-the-art quantum simulators from IBM Quantum and Xanadu using a variety of standard ML datasets. The experimental results showed favorable characteristics of our proposed techniques compared to existing work with respect to a number of quantitative metrics such as the number of training parameters, cross-entropy loss, classification accuracy, circuit depth, and quantum gate count.

Network Function Virtualization (NFV) allows for the dynamic provisioning of Virtualized Network Functions, adapting services to the complex and real-time network environment for enhanced network performance. Despite the benefits of NFV, virtual network functions (VNFs) migration and the energy consumption pose significant challenges due to the dynamic nature of the physical network, especially in the edge-cloud continuum network. Currently, people tackle this challenge by predicting the network traffic and then proactively migrating the VNFs using the predicted values. There are two challenges here. First, when SFCs belong to different network service providers, the historical network traffic data is held by the service providers, and they are reluctant to share the historical data due to privacy concerns. Without the historical network traffic data of each SFC, the resource provider that owns the underlying network cannot effectively predict the network traffic; in addition, without the feedback of the migration results of the underlying physical network, the network traffic prediction will be difficult to match the needs of the physical network. In this manuscript, we solve the problem of network traffic prediction in the NFV-based edge-cloud continuum network for multiple network service providers. We aim to improve migration performance and reduce energy consumption while ensuring resource and latency requirements. To protect the privacy of each network service provider, we apply an FL algorithm. After that, to improve the migration performance, we tailor the loss function. In the loss function, factors such as the number of migrations, the number of migration failures, and the energy consumption of the NFV-based edge-cloud continuum network are added. Additionally, to obtain the results of the SFC migration, we develop a mathematical model for the problem of multiple network service providers in the NFV-based network. The effectiveness of our algorithm is evaluated through extensive simulations, and the results show a significant reduction in the number of migrated nodes and energy consumption while improving the SFC acceptance ratio compared to state-of-the-art schemes which only use the difference between predicted and actual flows to define the loss function.

This paper proposes a new approach for multi-focus images fusion based on Region Mosaicing on Contrast Pyramids (REMCP). A density-based region growing method is developed to construct a focused region mask for multi-focus images. The segmented focused region mask is decomposed into a mask pyramid, which is then used for supervised region mosaicking on a contrast pyramid. In this way, the focus measurement and the continuity of focused regions are incorporated and the pixel level pyramid fusion is improved at the region level. Objective and subjective experiments show that the proposed REMCP is more robust to noise than compared algorithms and can fully preserves the focus information of the multi-focus images meanwhile reducing distortions of the fused images.

Glaucoma, a global leading cause of blindness, necessitates accurate segmentation of the optic disc (OD) and optic cup (OC) for effective screening. However, existing segmentation methods based on convolutional neural networks (CNNs) suffer from high computational complexity and long inference times. In this paper, we propose an end-to-end segmentation method for the OD and OC using a lightweight MobileNetv3 as the feature extraction network. Our approach incorporates boundary branches and an adversarial learning network to achieve multi-label segmentation of the OD and OC. We evaluate the proposed method on three publicly available datasets: Drishti-GS, RIM-ONE-r3, and REFUGE. The experimental results demonstrate segmentation accuracies of 0.974/0.900, 0.966/0.875, and 0.962/0.880 for the OD and OC on the respective datasets, while significantly reducing inference time.

To secure logic circuits against the illegal copying the insertion of the so-called key bit inputs using XOR/XNOR gates has been proposed in . The inserted gates modify (lock) the behavior of the original circuit and only the correct choice of the secret key bits guarantees the equivalent behaviour of the locked circuit to the original circuit. To break this kind of security, i.e. to find the secret key bits, there was a SAT solver based attack proposed in . The aim of this work is to experimentally verify this attack on several locked circuits with various number of primary inputs and outputs and with various number of gates. Except for the random placement of the key gates we evaluate also the positioning with chaining elimination. The obtained and presented results show that we were able to unlock 10 of 11 circuits within few minutes on a common PC.

With the integrated development of the Internet, wireless sensor technology, cloud computing, and mobile Internet, there has been a lot of attention given to research about and applications of the Internet of Things. A Wireless Sensor Network (WSN) is one of the important information technologies in the Internet of Things; it integrates multi-technology to detect and gather information in a network environment by mutual cooperation, using a variety of methods to process and analyze data, implement awareness, and perform tests. This paper mainly researches the localization algorithm of sensor nodes in a wireless sensor network. Firstly, a multi-granularity region partition is proposed to divide the location region. In the range-based method, the RSSI (Received Signal Strength indicator, RSSI) is used to estimate distance. The optimal RSSI value is computed by the Gaussian fitting method. Furthermore, a Voronoi diagram is characterized by the use of dividing region. Rach anchor node is regarded as the center of each region; the whole position region is divided into several regions and the sub-region of neighboring nodes is combined into triangles while the unknown node is locked in the ultimate area. Secondly, the multi-granularity regional division and Lagrange multiplier method are used to calculate the final coordinates. Because nodes are influenced by many factors in the practical application, two kinds of positioning methods are designed: the unknown node is in the positioning unit or not. When the unknown node is on the side of the positioning unit, we use the method of vector similarity. Moreover, we use the centroid algorithm to calculate the ultimate coordinates of unknown node. When the unknown node is not on the side of the positioning unit, we establish a Lagrange equation containing the constraint condition to calculate the first coordinates. Furthermore, we use the Taylor expansion formula to correct the coordinates of the unknown node. In addition, this localization method has been validated by establishing the real environment.

Reconstructing 3D indoor scenes from 2D images has always been an important task in computer vision and graphics applications. For indoor scenes, traditional 3D reconstruction methods suffer from the lack of surface details, poor reconstruction of large flat surface textures and areas with uneven lighting, and many falsely reconstructed floating debris noises in the reconstructed models. We add adaptive normal priors to the neural implicit reconstruction process to optimize the network, and improve the accuracy of volume density prediction by adding regularization terms to the neural radiation field to constrain the volume density obtained by weight distribution, and learn a smooth SDF surface from the network to obtain an explicit mesh model. Experiments show that the method proposed in this paper outperforms the state-of-the-art methods on ScanNet, Hypersim, and Replica datasets.

Domain generalization has led to remarkable achievements in Hyperspectral Image (HSI) classification. Inspired by contrastive language-image pre-training (CLIP), the language-aware domain generalization method has been explored to learn cross-domain-invariant representation. However, existing methods face some challenges: 1) The weak capacity to extract long-range contextual information and inter-class correlation. 2) Due to the inadequacies of the large-scale pre-training for HSI data, the spatial-spectral features of HSI and linguistic features can not be straightforwardly alignment. To address the above problems, a novel network has been proposed with a CLIP framework, which consists of an image encoder, based on an encoder-only transformer to obtain the global contextual information and inter-class correlation, a frozen text encoder, and a cross-attention mechanism, named Linguistic-Interact-with-Visual Engager (LIVE), enhances the interaction between two modalities. Extensive experiments demonstrating superior performance over state-of-the-art methods in HSI Domain Generalization with a CLIP framework.

Open-vocabulary learning has recently gained prominence as a means to enable image segmentation for arbitrary categories based on textual descriptions. This advancement has extended the applicability of segmentation systems to a broader range of generally purpose scenarios. However, current methods often revolve around specialized architectures and parameters tailored to specific segmentation tasks, resulting in a fragmented landscape of segmentation models. In response to these challenges, we introduce OVAMTSeg, a versatile framework designed for Open-Vocabulary and Multitask Image Segmentation. OVAMTSeg harnesses adaptive prompt learning to empower the model to capture category-sensitive concepts, enhancing its robustness across diverse multi-task and scenario contexts. Text prompts are employed to effectively capture semantic and contextual features of the text, while cross-attention and cross-modal interactions enable the fusion of image and text features. Furthermore, a transformer-based decoder is incorporated for dense prediction. Extensive experimental results underscore the effectiveness of OVAMTSeg, showcasing its state-of-the-art performance and superior generalization capabilities across three segmentation tasks. Notable achievements include a 47.5 mIoU in referring expression segmentation, 51.6 mIoU on Pascal-VOC with four unseen classes, 46.6 mIoU on Pascal-Context in zero-shot segmentation, 65.9 mIoU on Pascal-5i, and 35.7 mIoU on COCO-20i datasets for one-shot segmentation.

This project designed a set of network security mechanisms for cloud applications in VANETs. At present, cloud computing is one of the government’s development priorities in the industry. We divide cloud into public, private, and hybrid clouds in this project. Vehicles or passengers can access road conditions or public transport information through the public cloud; public transportation can access current driving records and users can access relevant enterprise information on the private cloud and hybrid cloud is a combination of public and private clouds. Both cloud computing or VANETs require network and information security. At present, many researches on VANETs security only focus on message communication, and neglect information storage. Researches on cloud computing security only focus on information protection, and neglect user privacy and anonymity. This project designed a set of network and information security mechanisms in line with the requirements of Confidentiality, Authentication, Non-repudiation, Conditional Anonymity, and Conditional Untraceability. This project primarily needs to achieve the following: 1.an authentication mechanism to verify the identity of each other between the passenger and the vehicle and to verify the identity with Single Sign-On; 2. vehicle or user privacy and anonymity, which need to be able to replace the anonymous ID and related parameters for the vehicle or the user; 3. a private communication mechanism, which enables any vehicle or user to communicate privately; and 4. an information security encryption method, which can encrypt the information on a cloud server to avoid unauthorized access by internal personnel or hackers.

The extent of correlation or co-movement among the returns of developed and/or emerging stock markets remains pivotal for efficiently diversifying global portfolios. This correlation is prone to variation over time as a consequence of escalating economic interdependence fostered by international trade and financial markets. In this study, we analyzed the time-varying correlation and co-movement between the JSE.JO stock market of South Africa and its developed and developing stock market partners. The Dynamic Conditional Correlation - Exponential Generalized Autoregressive Conditional Heteroscedasticity (DCC-EGARCH) methodology is employed with different multivariate distributions to explore the time-varying correlation and volatilities between the JSE.JO stock market and its partners. Based on the conditional correlation results, the JSE.JO stock market is integrated and co-moves with its partners, and the conditional correlation for all markets exhibit time-variant behavior. The conditional volatility results show that the JSE.JO stock market behaves differently from other markets, especially after 2015, indicating a positive sign for investors to diversify between JSE.JO and its partners. The highest value of conditional volatility for markets was in 2020 during the COVID-19 pandemic, representing the riskiest period that investors should avoid due to the lack of diversification opportunities during crises.

This article aims to explore the existence and stability of solutions to differential equations involving a regularized ψ−Hilfer fractional derivative, which, compared to standard ψ−Hilfer fractional derivatives, provide a clear physical interpretation when dealing with initial conditions. We discovered that the regularized ψ−Hilfer fractional derivative can be represented as the inverse operation of the ψ−Riemann–Liouville fractional integral, and used this property to prove the existence of solutions for linear differential equations with a regularized ψ−Hilfer fractional derivative. Additionally, we applied the Mönch’s fixed-point theorem and knowledge of non-compactness measures to demonstrate the existence of solutions for nonlinear differential equations with a regularized ψ−Hilfer fractional derivative, and further discussed Ulam-Hyers-Rassias stability and semi-Ulam-Hyers-Rassias stability of these solutions. Finally, we illustrated our results through case studies.

The aim of the present study is to investigate both the performance and preferences of males and females Computer Science (CS) graduates. In order to attain the above goal, a quantitative case study was conducted regarding 89 degrees, acquired from 2006 to 2012, from the Department of Computer Science and Technology, University of Peloponnese, Greece. The analysis of the data revealed that in terms of performance, no significant differences between the mean grades of males and females exist, in almost most of the courses included in the curriculum of the aforementioned CS department. Any statistically significant differences in performances were present in almost equal number of courses in favor of males and females. It seems also, that females performed better in the courses they selected more than males. Regarding preferences, in CS courses, it seems that gender differences are existent. Males preferred more than females did core programming courses and advanced topics of Software Systems, computer networks, computer engineering, robotics and mathematics, whereas females preferred more the study of algorithms and security issues, computer fractals, data management, computer architecture, and mobile communication. In addition, females preferred courses in reference with humanities and social sciences, CS terminology, and career opportunities. Yet, females did not select any of programming lab-based courses, computer engineering, computer network issues and robotics.

Fire detection is crucial for ensuring human safety and minimizing property damage. By utilizing advanced technologies, we can identify fires early, before they escalate. Autonomous fire detection systems are particularly vital in high-risk areas with minimal human presence. To address this need, we propose an automated fire-detection model based on YOLOv8. We modified the architecture of YOLOv8 and achieved impressive performances. Through comprehensive analysis of different YOLO models and their predecessors, we identified and addressed substantial gaps in this field. Also, by augmenting the framework with additional capabilities, we surpassed previous models, enabling real-time fire detection. Our proposed model achieves impressive performances, with 98% accuracy in fire detection and 97.8% in smoke detection. Moreover, recall rates for fire and smoke are 97.1% and 97.4%, respectively, with a mean average precision (mAP) accuracy of 99.1% for both cases. Finally, we applied an explainable artificial intelligence technique to interpret our proposed model and its results. This study lays the foundation for future research aimed at enhancing fire detection efficiency. The reliability of our modified model presents promising opportunities for future advancements, ensuring increased effectiveness and precision in efficient fire detection.

We propose an innovative approach called SMPTC3, designed specifically to enhance security and privacy in cross-chain transaction verification. This approach addresses multi-chain, multi-participant information exchange and large-scale cross-chain transfers, resisting various types of malicious attacks. We leverage the verifiability of cross-chain transactions based on smart contracts and innovatively transform transaction information into confidential sets, organizing them into quadratic secret polynomials. By utilizing secret sharing and random distribution techniques, we construct a secure multiparty computation method, tailored for cross-chain transactions. To enhance the efficiency of cross-chain transactions, we introduce cross-chain batch processing technology, grouping inter-chain transactions into cross-chain transaction sets. Unlike traditional distributed notary technologies, SMPTC3 designates honest participants as a cross-chain notary group, reducing the time required for redundant signature confirmations and significantly lowering the possibility of malicious notaries. Theoretical analysis and empirical experiments demonstrate that SMPTC3 is highly efficient in addressing cross-chain transaction security issues.

Detecting cracks in structural health monitoring is crucial for ensuring infrastructure safety and longevity. Using drones to obtain crack images and automate processing can improve the efficiency of crack detection. To address the challenges posed by the limited computing resources of edge devices in practical applications, we propose CrackScopeNet, a lightweight segmentation network model that simultaneously considers local and global crack features, suitable for deployment on drone platforms with limited computational power and memory. CrackScopeNet features a multi-scale branch to improve sensitivity to cracks of varying sizes without substantial computational overhead and a stripe-wise context attention mechanism to enhance the capture of long-range contextual information, mitigating the interference from complex backgrounds. Experimental results on CrackSeg9k dataset demonstrate that CrackScopeNet leads to a significant improvement in prediction performance, with the highest mean intersection over union (mIoU) scores reaching 82.12%, while maintaining a lightweight architecture with only 1.05M parameters and 1.58G floating point operations (FLOPs). Besides, CrackScopeNet excels in inference speed on edge devices without GPU because of its low FLOPs. CrackScopeNet contributes to the development of efficient and effective crack segmentation networks, suitable for practical structural health monitoring applications using drone platforms.

The spread of fake news poses a serious threat to public trust and informed decision-making. The emergence of AI-generated fake news exacerbates this issue, as it can be produced more rapidly and convincingly than human-generated content. Addressing this issue, our paper introduces a robust approach to detect AI-generated fake news that is applicable across major disinformation domains: Politics and Elections, Health and Medicine, Science and Technology, and Entertainment and Celebrities. Our models differentiate between AI-generated and human-generated news articles while assessing their truthfulness. Utilizing large language models such as ChatGPT, we curated datasets spanning both AI and human generated fake and genuine news. Following extensive data preprocessing and the application of diverse feature extraction techniques, our models were trained using multiple machine learning algorithms. Our results show the feasibility of accurately identifying AI-generated fake news, thereby adding a significant layer of defense in combating misinformation across various sectors. In addition, we have explored the potential utility of transferable machine learning models trained on data from one domain (e.g., politics), when tested on another domain (e.g., health).

In the extended malfatti problem, the numbers of circles in the triangle are triangle numbers, because the tangency properties among the internal circles and 3 sides of the triangle have special type of structure. That is, the corner circle tangent to two sides of the triangle and two boundary circles, the boundary circles tangent to one side of the triangle and 4 other circles, and the inner circles always tangent to 6 other circles. The circles we find in extended general Malfatti’s problem has the property that the smallest radius and the largest radius for the circle has big difference. In this paper, we proposed algorithms to solve the problem that the tangency properties among circles and sides of triangle is not fixed, so that the number of circles in the triangle is not necessary a triangle numbers. The purpose of this change is try to find the radii of the circles in the triangle is within a small range.

Survivability is a critical concern in WSNs, heavily influenced by energy efficiency. Addressing severe energy constraints in WSNs requires solutions that meet application goals while prolonging network life. This paper presents an energy optimization approach (EOAMRCL) for WSNs, integrating the Grey Wolf Optimization (GWO) for enhanced performance. EOAMRCL aims to enhance energy efficiency by selecting the optimal duty-cycle schedule, transmission power, and routing paths. The proposed approach employs a centralized strategy using a hierarchical network architecture. During the cluster formation phase, an objective function, augmented with GWO, determines the ideal cluster heads (CHs). The routing protocol then selects routes with minimal energy consumption for data transmission to CHs, using transmission power as a metric. In the transmission phase, the MAC layer forms a duty-cycle schedule based on cross-layer routing information, enabling nodes to switch between active and sleep modes according to their network allocation vectors (NAV). This process is further optimized by an enhanced CSMA/CA mechanism, which incorporates sleep/activate modes and pairing nodes to alternate between active and sleep states. This integration reduces collisions, improves channel assessment accuracy, and lowers energy consumption, thereby enhancing overall network performance. EOAMRCL was evaluated in a MATLAB environment, demonstrating superior performance compared to EEUC, DWEHC, and CGA-GWO protocols, particularly in terms of network lifetime and energy consumption. This highlights the effectiveness of integrating GWO and the updated CSMA/CA mechanism in achieving optimal energy efficiency and network performance.

The open data market is growing, driven by laws requiring governments to make data publicly available. However, the true value of this market lies in the products and services derived from open data. In Slovenia, the open data market has been growing since 2015, especially in the field of mobility. Platforms such as Avtolog.si facilitate the retrieval of vehicle data via vehicle identification numbers, while tocen.si, ojpp.si, and brezavta.si aggregate mobility data, focusing on sustainable and micromobile public transportation services. These platforms use open government data and private sector open data, and are freely accessible. This study examines the use of open data in the Slovenian mobility applications and similar solutions developed in other countries. By mapping these applications, we gain valuable insights into potential advances in the field, and provide a useful resource for future developers worldwide. By focusing on the socio-economic impact of these innovations, the study identifies methods for evaluating such applications. This approach highlights how transforming open data into user-friendly products and services can deepen our understanding of their broader societal and economic benefits. While open data is valuable in its own right, its full transformative potential is realized through innovative products that support sustainable lifestyles and promote environmentally conscious behavior.

Despite significant advancements in Artificial Intelligence (AI) and Large Language Models (LLMs), detecting and mitigating bias remains a critical challenge, particularly within social media platforms like X (formerly Twitter) and in addressing cyberbullying present on them. This research investigates the effectiveness of leading LLMs in generating synthetic biased and cyberbullying data and evaluates the proficiency of Transformer AI models in detecting bias and cyberbullying within both authentic and synthetic contexts. The study involves semantic analysis and feature engineering on a dataset of over 48,000 sentences related to cyberbullying collected from Twitter (before it became X). Leveraging state-of-the-art LLMs such as ChatGPT-4o, Pi AI, Claude 3 Opus, and Gemini-1.5, synthetic biased, cyberbullying, and neutral data were generated to deepen the understanding of bias in human-generated data. AI models including DeBERTa, Longformer, BigBird, HateBERT, MobileBERT, DistilBERT, BERT, RoBERTa, ELECTRA, and XLNet were initially trained to classify Twitter cyberbullying data and subsequently fine-tuned, optimized, and quantized for multilabel classification (detecting both biases and cyberbullying). The study's outcomes include a prototype of a hybrid application that combines a Bias Data Detector and a Bias Data Generator, validated through extensive testing.

With increasing research focus on industrial anomaly detection, numerous methods have emerged in this domain. Notably, memory bank-based approaches, coupled with kNN distance metrics, have demonstrated remarkable performance in anomaly detection (AD) and anomaly segmentation (AS). However, upon examination of the Back to the Feature (BTF) method applied to the Mvtec-3D AD dataset, it was observed that while it exhibited exceptional segmentation performance, its detection performance was lacking. To address this discrepancy, this study base improves the implementation of BTF, especially the improvement of the anomaly score metric. It posits that different "clusters" necessitate distinct k values in kNN distance metrics. For simplify, this assumption is distilled into the proposition that AD and AS tasks impose differing requirements on the k value in kNN distance metrics. Consequently, the paper introduces the BTM method, which utilizes distinct distance metrics for AD and AS tasks. This innovative approach yields superior AD and AS performance (I-AUROC 93.0%, AURPO 96.9%, P-AUROC 99.5%), representing a substantial enhancement over the BTF method (I-AUROC 5.7% ↑, AURPO 0.5% ↑, P-AUROC 0.2% ↑).

One of the most prevalent illnesses, diabetes does not directly result in patient mortality. But, it increases the risk of death. Any disease that may be predicted in its early stages can lessen its fatal effects while also enhancing the quality of the healthcare system. For the early-stage prediction of diabetes or such types of non-communicable diseases, we need the proper set of influential features. This research has developed a machine learning-based disease prediction model to identify the influential features for diabetes prediction and give a near-perfect classification accuracy. This model includes Min-Max normalization for data normalization, Isolation Forest (iForest) for outlier removal, Synthetic Minority Oversampling Technique (SMOTE) for oversampling, Random Forest based Recursive Feature Elimination (RFE-RF) test, Chi-Square test, and Minimum Redundancy Maximum Relevancy (mRMR) test based feature selection methods for identifying the influential features, and Support Vector Machine (SVM), K Nearest Neighbor (KNN), and Naive Bayes (NB) for the classification. The results clarify that the proposed model outperforms the previous models and studies. The SVM has attained an accuracy of 99.58% in classification using the five features chosen from the Chi-Square test. Lastly, SHAP, an explainable AI model, has been used to assess the classifier model’s performance. These selected features and the classifier model can be used for early-stage diabetes prediction.

The paper examines the various uncertainties encountered in high-frequency trading (HFT) environments and delves into the multiple challenges faced by HFT firms in navigating the Dodd-Frank Wall Street Reform and Consumer Protection Act (referred to as the "Dodd-Frank Act"), particularly during the initial stages of its enactment. These challenges include the ambiguity surrounding the definition of HFT, the lack of clarity regarding regulatory requirements and boundaries, inconsistencies in enforcement resulting from deviations in understanding the content, and the absence of detailed descriptions of the Act’s provisions. These hurdles significantly impact not only the daily operations of HFT firms but also pose higher demands on their long-term strategic planning and risk management. Drawing upon the Data, Information, Knowledge, Wisdom, and Purpose (DIKWP) model, this study employs an innovative analytical framework. Through the comprehensive application of conceptual space, cognitive space, and semantic space, it provides a systematic methodology for identifying and analyzing the aforementioned issues. This approach not only aids firms in better comprehending and adhering to complex regulatory requirements but also enables them to explore new business opportunities and competitive advantages while ensuring compliance.

End-to-end learned image compression codecs have notably emerged in recent years. These codecs have demonstrated superiority over conventional methods, showcasing remarkable flexibility and adaptability across diverse data domains while supporting new distortion losses. Despite challenges such as computational complexity, learned image compression methods inherently align with learning-based data processing and analytic pipelines due to their well-suited internal representations. The concept of Video Coding for Machines has garnered significant attention from both academic researchers and industry practitioners. This concept reflects the growing need to integrate data compression with computer vision applications. In light of these developments, we present a comprehensive survey and review of lossy image compression methods. Additionally, we provide a concise overview of two prominent international standards, MPEG Video Coding for Machines and JPEG AI. These standards are designed to bridge the gap between data compression and computer vision, catering to practical industry use cases.

The exceptional performance advantages of the fiber optic gyroscope (FOG) position it as a dominant player in middle and high-end inertial navigation systems. To prevent the loss of sensor precision caused by algorithm design and simplify the complex modeling strategy in traditional methods. We gradually demonstrate the significant role of Convolutional Neural Network (CNN) in the navigation system based on FOG, and utilize the particle swarm optimization algorithm (PSO) to expedite the convergence of the network. The experimental results demonstrate that the initial alignment method based on deep learning is more accurate than the traditional method. The attitude angle error is reduced by 81.25%, 92.54% and 36.53% respectively. The research provides support for the future application of deep learning in optical navigation systems.

Cloud Computing is a promising and emerging technology that is rapidly being adopted by many IT companies due to a number of benefits that it provides, such as large storage space, low investment cost, virtualization, resource sharing, etc. Users are able to store a vast amount of data and information in the cloud and access it from anywhere, anytime on a pay-per-use basis. Since many users are able to share the data and the resources stored in the cloud, there arises a need to provide access to the data to only those users who are authorized to access it. This can be done through access control schemes which allow the authenticated and authorized users to access the data and deny access to unauthorized users. In this paper, a comprehensive review of all the existing access control schemes has been discussed along with analysis. Keywords: role-based access control, attribute-based access control, attribute-based encryption

"Public legal information from all countries and international institutions is part of the common heritage of humanity. Maximizing access to this information promotes justice and the rule of law". In accordance with the aforementioned declaration on Free Access to Law by Legal information institutes of the world, a plethora of legal information is available through the Internet, while the provision of legal information has never before been easier. Given that law is accessed by a much wider group of people, the majority of whom are not legally trained or qualified, diversification techniques, should be employed in the context of legal information retrieval, as to increase user satisfaction. We address diversification of results in legal search by adopting several state of the art methods from the web search, network analysis and text summarization domains. We provide an exhaustive evaluation of the methods, using a standard data set from the Common Law domain that we subjectively annotated with relevance judgments for this purpose. Our results i) reveal that users receive broader insights across the results they get from a legal information retrieval system, ii) demonstrate that web search diversification techniques outperform other approaches (e.g., summarization-based, graph-based methods) in the context of legal diversification and iii) offer balance boundaries between reinforcing relevant documents or sampling the information space around the legal query.

Modular inversion on large operands is a time-consuming calculation used in elliptic curve cryptography. Its hardware implementation requires extensive hardware resources such as lookup tables and registers. We investigate state-of-the-art modular inversion algorithms and evaluate the performance and cost of the algorithms and their hardware implementations. We then propose a high-radix modular inversion algorithm aimed at short execution time and low hardware cost. We present a detailed radix-8 hardware implementation based on 256-bit primes in Verilog HDL and compare its cost performance with other implementations. Our implementation on the Altera Cyclone V FPGA chip uses 1227 ALMs (Adaptive logic modules) and 1037 registers. The modular inversion calculation takes 3.67 microseconds. The AT (Area time) factor is 8.30, outperforming other implementations. We also present an implementation of elliptic curve cryptography using the proposed radix-8 modular inversion algorithm. The implementation results also show that our modular inversion algorithm is more efficient in area time than other algorithms.

Sharding shows great potential for extending the blockchain. The current challenge facing sharded blockchain technology lies in addressing the extended transaction confirmation times caused by isolated states between shards and unbalanced transaction allocation strategies. These factors contribute to an increase in cross-shard transactions and disproportionate shard workload, ultimately resulting in indefinite confirmation delays for cross-shard transactions. A critical priority for sharded blockchain systems is to conduct a comprehensive qualitative analysis to better understand and mitigate the prolonged transaction confirmation times. We introduce a Product-Form Queue Network (PFQN) model to address the transaction confirmation time problem in sharded blockchains and incorporate a new confirmation queue to more accurately simulate the actual transaction confirmation process in the blockchain. In addition, we provide a detailed quantitative analysis of the relationship between network load, consensus efficiency, and security in sharded blockchains, laying a theoretical foundation for achieving robustness and efficiency in sharded blockchains. This research not only contributes to addressing the scalability issues in sharded blockchains but also offers a new perspective for future research directions.

As the adoption of deep learning models continues to surge across various applications, the need for efficient deployment architectures becomes increasingly critical. This paper presents a novel approach to enhance the deployment of deep learning models by leveraging serverless architecture. Serverless computing has been popular for its auto-scaling, cost-effectiveness, and simplified management characteristics. However, the intense resource demands of deep learning models pose challenges in maintaining low response times and effective load balancing within serverless environments. The proposed architecture addresses these challenges by integrating principles from both deep learning model optimization and serverless computing. Through systematic experimentation and analysis, we demonstrate that by appropriately designing and tuning the deployment architecture, significant improvements in response time, performance, resource utilization, and load distribution can be achieved.

The applicability of the distance aggregation problem has attracted the interest of many authors. Motivated by this fact, in this paper we face the modular quasi-(pseudo-)metric aggregation problem. We characterize those functions that allow merging a collection of modular quasi-(pseudo-)metrics into a single one. Specifically, a description of such functions in terms of triangle triplets is given and, in addition, the relationship between modular quasi-(pseudo-)metric aggregation functions and modular (pseudo-)metric aggregation functions is discussed. Such characterizations are illustrated with appropriate examples. A few methods to construct modular quasi-(pseudo-)metrics are yielded. Several properties of modular quasi-(pseudo-)metric aggregation functions are explored and used to develop quick tests for discarding candidate functions to aggregate modular quasi-(pseudo-)metrics. Moreover, a characterization of those modular quasi-(pseudo-)metric aggregation functions that preserve modular quasi-(pseudo-)metrics is also provided. Furthermore, the relationship between modular quasi-(pseudo-)metric aggregation functions and quasi-(pseudo-)metric aggregation functions is studied in such a way that significative differences are displayed.

This research introduces a hybrid feature extraction approach that combines Linear Discriminant Analysis (LDA) and Multilayer Perceptron (MLP) methods to address the challenges of reducing feature vector dimensionality and accurately classifying smartphone-based human activities. Moreover, to refine activity classification accuracy, Support Vector Machine (SVM) optimization with Stochastic Gradient Descent (SGD) is employed. LDA, a statistical tool, is leveraged to derive a new feature space for data projection, enhancing class separation and test feature label prediction. The proposed approach, named LMSS, was evaluated using the UCI-HAR dataset and compared with state-of-the-art models. The results demonstrate that the proposed approach outperformed the best-performing method over this dataset. It achieved an accuracy rate of 99.52%, precision of 99.55%, recall of 99.53%, and an F1-score of 99.54%, highlighting the effectiveness of the proposed method in accurately classifying the data.

Dead reckoning is a promising yet often overlooked smartphone-based indoor localization technology that relies on phone-mounted sensors for counting steps and estimating walking directions, without the need for extensive sensor or landmark deployment. However, misalignment between the phone’s direction and the user’s actual movement direction can lead to unreliable direction estimates and inaccurate location tracking. To address this issue, this paper introduces SWiLoc, an enhanced direction correction system that integrates passive WiFi sensing with smartphone-based sensing to form Correction Zones. Our two-phase approach accurately measures the user’s walking directions when passing through a Correction Zone and further refines successive direction estimates outside the zones, enabling continuous and reliable tracking. In addition to direction correction, SWiLoc extends its capabilities by incorporating a localization technique that leverages corrected directions to achieve precise user localization. This extension significantly enhances the system’s applicability for high-accuracy localization tasks. Additionally, our innovative Fresnel zone-based approach, which utilizes unique hardware configurations and a fundamental geometric model, ensures accurate and robust direction estimation, even in scenarios with unreliable walking directions. We evaluate SWiLoc across two real-world environments, assessing its performance under varying conditions such as environmental changes, phone orientations, walking directions, and distances. Our comprehensive experiments demonstrate that SWiLoc achieves an average 75th percentile error of 8.89 degrees in walking direction estimation and an 80th percentile error of 1.12 meters in location estimation. These figures represent reductions of 64% and 49%, respectively for direction and location estimation error, over existing state-of-the-art approaches.

Over the last decade, a lot of research has been conducted on memristors. Most of this research focusses on using memristors for Artificial Intelligence (AI) applications and to fabricate non-volatile memory, but also the security aspects of memristors have been examined. The current study summarises and compares five reviews on the security aspects of memristors. These reviews cover two different perspectives: (1) security applications of memristors such as Physical Unclonable Functions (PUFs) or True Random Number Generators (TRNGs), and (2) potential threats when using memristors to train and store neural networks. The comparison of the reviews reveals that different sets of studies are included in the reviews and different characterisations of the studies are provided. This shows that different perspectives are necessary to get a comprehensive overview of the security aspects of memristors. By synthesising the perspectives of different reviews, this study helps to get such an overview.

In the dynamic landscape of modern education, characterized by an increasingly active involvement of IT technologies in learning, the imperative to transfer to students the skills necessary to integrate Artificial Intelligence (AI) into the process represents an important goal. This paper presents a novel framework for knowledge transfer, diverging from traditional programming language-centric approaches by integrating PSoC 6 microcontroller technology. This framework establishes a cyclical learning cycle encompassing theoretical fundamentals and practical experimentation, fostering AI literacy at the edge. Through a structured combination of theoretical instruction and hands-on experimentation, students develop proficiency in understanding and harnessing AI capabilities. Emphasizing critical thinking, problem-solving, and creativity, this approach equips students with the tools to navigate the complexities of real-world AI applications effectively. By leveraging PSoC 6 as an educational tool, a new generation of individuals is efficiently cultivated with essential AI skills. These individuals are adept at leveraging AI technologies to address societal challenges and drive innovation, thereby contributing to long-term sustainability initiatives. Specific strategies for experiential learning, curriculum recommendations, and results of knowledge application are presented, aimed at preparing students to excel in a future where AI will be omnipresent and indispensable.

In response to the issues of sensor noise and accumulated motion errors that lead to low positioning accuracy and insufficient robustness during large maneuvers or continuous flight of small unmanned aerial vehicles (UAVs) in complex environments, this paper proposes a multi-source and multimodal data fusion method. Initially, it employs a multimodal data fusion of various sensors including GPS (Global Positioning System), IMU (Inertial Measurement Unit), and visual sensors to complement the strengths and weaknesses of each hardware component, thereby eliminating motion errors to enhance accuracy. To mitigate the impact of sudden changes in sensor data, the method introduces a digital twin UAV data source. By utilizing the Extended Kalman Filter algorithm, it fuses data from both the real UAV and its digital twin, and the filtered positional information is fed back into the control system of the real UAV. This enables real-time correction of UAV positional deviations caused by sensor noise and environmental disturbances. The multi-source multimodal fusion Kalman filter method proposed in this paper significantly improves the positioning accuracy of UAVs in complex scenarios and the overall stability of the system. This method holds significant value in maintaining high-precision positioning in variable environments and has important practical implications for enhancing UAV navigation and application efficiency.

Deploying Cellular Internet of Things (CIoT) devices in urban and remote areas faces significant energy efficiency challenges. This is especially true for Narrowband IoT (NB-IoT) devices, which are expected to function on a single charge for up to 10 years while transmitting small amounts of data daily. The 3rd Generation Partnership Project (3GPP) has introduced energy-saving mechanisms in Releases 13 to 16, including Early Data Transmission (EDT) and Preconfigured Uplink Resources (PUR). These mechanisms extend battery life and reduce latency by enabling data transmission without an active Radio Resource Control (RRC) connection or Random Access Procedure (RAP). This paper examines these mechanisms using the LENA-NB simulator in the ns-3 environment, which is a comprehensive framework for studying various aspects of NB-IoT. The LENA-NB has been extended with PUR, and our analysis shows that PUR significantly enhances battery life and latency efficiency, particularly in high-density environments. Compared to the default RAP method, PUR reduces energy consumption by more than 2.5 times and increases battery life by 1.6 times. Additionally, PUR achieves latency reductions of 2.5-3.5 times. The improvements with PUR are most notable for packets up to 125 bytes. Our findings highlight PUR’s potential to enable more efficient and effective CIoT deployments across various scenarios. This study represents a detailed analysis of latency and energy consumption in a simulated environment, advancing the understanding of PUR’s benefits.

Chemical vapor deposition (CVD) is a vital process for depositing thin films of various materials with precise control over thickness, composition, and properties. Understanding the heat and mass transfer mechanisms during CVD is essential for optimizing process parameters and ensuring high-quality film deposition. This review provides an overview of recent advancements in the heat and mass transfer modeling for chemical vapor deposition processes. It explores innovative modeling techniques, recent research findings, emerging applications, and challenges in the field. Additionally, it discusses future directions and potential areas for further advancement in CVD modeling.

This article aims to delve into the Equity, Diversity and Inclusivity (EDI) issues prevalent in the engineering disciplines in Canada and Spain, shedding light on the common obstacles faced by underrepresented individuals and highlighting potential strategies to foster a more inclusive and diverse engineering community in these nations. Two strategic lines have been identified: (a) facilitating university education access to underrepresented and minority groups, and (b) Accompanying and guiding such students during university training, setting them up for successful future careers. The article also shows the sets of strategies employed in Canada and Spain, clearly distinguishing the approach taken in the two countries. While in Canada, there is a more decentralized approach wherein the universities device their strategies and agenda to address EDI issues, in Spain there is a stronger and direct involvement of the government to ensure a comprehensive, system-wide approach to tackling EDI issues in academia.

Device to Device (D2D) is a pivotal technology in the next generation of communication, allowing for direct task offloading between mobile devices (MDs) to improve the efficient utilization of idle resources. This paper proposes a novel algorithm for dynamic task offloading between active MDs and idle MDs in the D2D-MEC (Mobile Edge Computing) system by deploying multi-agent deep reinforcement learning (DRL) to minimize the long-term average delay of delay-sensitive tasks under deadline constraints. Our core innovation is a dynamic partitioning scheme for idle and active devices in the D2D-MEC system, accounting for stochastic task arrivals and multi-time-slot task execution, which has been insufficiently addressed in existing literature. We adopt a queue-based system to formulate a dynamic task offloading optimization problem. To address the challenges of large action space and coupling of actions across time slots, we model the problem as a Markov Decision Process (MDP) and perform the multi-agent DRL through Multi-Agent Proximal Policy Optimization (MAPPO). We employ a centralized training with decentralized execution (CTDE) framework to enable each MD to make offloading decisions solely based on its local system state. Extensive simulations demonstrate the efficiency and fast convergence of our algorithm. In comparison with the existing sub-optimal results deploying single-agent DRL, our algorithm reduces the average task completion delay by 11.0% and the ratio of dropped tasks by 17.0%.

Dynamic sparse training is an effective strategy to alleviate the training and inference demands of artificial neural networks. However, current sparse training methods face a challenge in achieving high levels of sparsity while maintaining performance comparable to that of their fully connected counterparts. The Cannistraci-Hebb training (CHT) method produces an ultra-sparse advantage compared to fully connected training in various tasks by using a gradient-free link regrowth method, which relies solely on the network topology. However, its rigid selection based on link prediction scores may lead to epitopological local minima, especially at the beginning of the training process when the network topology might be noisy and unreliable. In this article, we introduce the Cannistraci-Hebb training soft rule (CHTs), which applies a flexible approach to both the removal and regrowth of links during training, fostering a balance between exploring and exploiting network topology. Additionally, we investigate the network topology initialization using several approaches, including the bipartite scale-free and bipartite small-world network models. Empirical results show that CHTs can surpass the performance of fully connected networks with MLP architecture by using only 1% of the connections (99% sparsity) on the MNIST, EMNIST, and Fashion MNIST datasets and can provide remarkable results with only 0.1% of the links (99.9% sparsity). In some MLPs for image classification tasks, CHTs can reduce the active neuron network size to 20% of the original nodes (neurons), demonstrating a remarkable ability to generalize better than fully connected architectures, reducing the entire model size. This represents a relevant result for dynamic sparse training. Finally, we present evidence from larger network models such as Transformers, with 10% of the connections (90% sparsity), where CHTs outperform other prevalent dynamic sparse training methods in machine translation tasks.

Traditional 3D reconstruction models have consistently encountered a challenge: attaining high recall of object edges while preserving precision. In this paper, we introduce a post-processing method Multi-Head Attention Refiner (MA-R) aimed at tackling this challenge by integrating a multi-head attention mechanism within the U-Net style refiner module. The 3D reconstruction model applying this method demonstrates excellence in parsing intricate image details, resulting in significant enhancement of boundary predictions and increased recall rates. In our experiments, our method significantly enhances the reconstruction performance of Pix2Vox++ when multiple images are utilized as input. Specifically, it achieves an enhanced Intersection Over Union score of approximately 1.1% compared to the original Pix2Vox++ model when 16-view images are employed.

Pedestrian trajectories are crucial for self-driving cars to plan their paths effectively. The sensors implanted in these self-driving vehicles, despite being state-of-the-art ones, often face inaccuracies in the perception of surrounding environments due to technical challenges for adverse weather conditions, interference from other vehicles sensors and electronic devices, and signal reception failure, leading to incompleteness in trajectory data. But for real-time decision making for autonomous driving, the trajectory imputation is no less crucial. Previous attempts to address this issue such as statistical inference and machine learning approaches have shown promise. Yet, the landscape of deep learning is rapidly evolving with new and more robust models emerging. In this research, we have proposed an encoder-decoder architecture, Human Trajectory Imputation Model, coined as HTIM, to tackle these challenges. This architecture aims to fill in the missing parts of pedestrian trajectories. The model has been evaluated using the Intersection drone inD dataset, containing trajectory data at suitable altitudes preserving naturalistic pedestrian behavior with varied dataset sizes. To assess the effectiveness of our model, we have utilized L1, MSE, Quantile and ADE loss. Our experiments have demonstrated that HTIM outperforms the majority of the state-of-the-art methods in this field, thus indicating its superior performance in imputing pedestrian trajectories.

Feature selection is predominantly used in machine learning tasks, such as classification and regression. It selects a subset of features (relevant attributes of data points) from a larger set that contributes as optimally as possible to the informativeness of the model. There are exponentially many subsets of a given set, and thus, the exhaustive search is only practical for problems with at most a few dozen features. In the past, there have been attempts to reduce the search space using dynamic programming. However, models that consider similarity in pairs of features besides the quality of individual features do not provide the required optimal substructure. As a result, algorithms, which we will call suboptimal dynamic programming algorithms, find a solution that may deviate significantly from the optimal one. In this paper, we propose a dynamic programming algorithm with an iterative upgrade where the order of feature processing is inverted in each iteration. Such an alternating approach allows for improving the optimization function by using the score from the previous iteration to estimate the contribution of unprocessed features. The iterative process is proven to converge and terminate when the solution does not change in three successive iterations or when the number of iterations reaches the threshold. Results in more than 90 % of tests align with those of the exhaustive search, being competitive and often superior to the reference greedy approach. The validation was done by comparing the scores of output feature subsets and examining the accuracy of different classifiers learned on these features. In the context of feature selection, the proposed algorithm can be characterized as a filter method. However, we expect that the idea of alternating suboptimal optimization will soon be generalized to tasks beyond feature selection.

Backscatter communication, which transmits information by passively reflecting radio frequency (RF) signals, has become a focal point of interest due to its potential to significantly enhance the energy efficiency of Wireless Power Mobile Edge Computing (WPMEC) networks and extend the operational lifespan of terminal devices. However, there is little research on the integration of user cooperation in WPMEC scenarios within volatile network environments. In this paper, we propose a dynamic task offloading scheme for a Backscatter-assisted WPMEC system, which involves two mobile devices (MDs) and a Hybrid Access Point (HAP) with user cooperation. We formulate the energy efficiency (EE) maximization problem as a stochastic programming problem, considering the randomness of task arrivals and time-varying wireless channels. By leveraging Dinkelbach’s method and stochastic network optimization technique, we transform the problem into a series of deterministic sub-problems for each time slot, and convert the non-convex sub-problem into convex ones. We propose a low-complex EE maximization algorithm to solve the convex problems efficiently. Extensive simulations are conducted to validate the performance of our algorithm under various system parameter settings. Experiment results demonstrate that our algorithm not only outperforms the benchmark algorithms by approximately 23%, but also stabilize all queues within the MEC system.

Motion related human activity recognition using wearable sensors can potentially enable various useful daily applications. So far, most studies view it as a stand-alone mathematical classification problem without considering the physical nature of human motions. Consequently, they suffer from data dependencies and encounter the dimension disaster problem and the over-fitting issue, and their models are never human-readable. In this study, we start from a deep analysis on natural physical properties of human motions, and then propose a useful feature selection method to quantify each feature's classification contribution capability. On one hand, the "dimension disaster" problem can be avoid to some extent, due to the affined dimension of key features; On the other hand, over-fitting issue can be depressed since the knowledge implied in human motions are nearly invariant, which compensates the possible data inadequacy. The experiment results indicate that the proposed method performs superior to those adopted in related works, such as decision tree, k-NN, SVM, neural networks.

All types of sports are potential application scenarios for automatic and real-time visual object and event detection. In rink hockey, the popular roller quad skate variant of hockey team sports, it is of great interest to automatically track player’s movements and positions, player’s sticks and, also, making other judgments, such as being able to locate the ball. In this work, we introduce a real-time pipeline composed by an object detection model, created specifically for rink hockey games, followed by a knowledge-based event detection module. Even in the presence of occlusions and quick motions, our deep learning object detection model effectively identifies and tracks, in real-time, important visual elements such as: ball; players; sticks; referees; crowd; goalkeeper; and goal. Using a curated dataset composed by a collection of videos of rink hockey, comprising 2525 annotated frames, we trained and evaluated the algorithm performance and compare it to state of the art object detection techniques. Our object detection model, based on YOLOv7, presents a global accuracy of 80%, and presents a good performance in terms of accuracy and speed, according to our results, making it a good choice for rink hockey applications. In our initial tests, the event detection module successfully detected one important event type in rink hockey games, the occurrence of penalties.

Gastric cancer is the fifth most common and fourth deadliest cancer worldwide, with a bleak 5-year survival rate of about 20%. Despite significant research into its pathobiology, prognostic predictability remains insufficient due to pathologists’ heavy workloads and the potential for diagnostic errors. Consequently, there is a pressing need for automated and precise histopathological diagnostic tools. This study leverages Machine Learning and Deep Learning techniques to classify histopathological images into healthy and cancerous categories. By utilizing both handcrafted and deep features and shallow learning classifiers on the GasHisSDB dataset, we conduct a comparative analysis to identify the most effective combinations of features and classifiers for differentiating normal from abnormal histopathological images without employing fine-tuning strategies. Our methodology achieves an accuracy of 95% with the SVM classifier, underscoring the effectiveness of feature fusion strategies. Additionally, cross-magnification experiments produced promising results with accuracies close to 80% and 90% when testing the models on unseen testing images with different resolutions.

Abstract: Author attribution refers to identifying the author of a document by distinguishing between texts written by various authors. Since the author’s writing style is a straightforward concept of his writing habits, an author’s attribution can be used in various sections of society, including legal work, court cases, and plagiarism. This paper studies and compares a deep architecture that automatically identifies Persian writers through their writing styles. For this purpose, since there were no existing datasets suitable for author identification in Persian, the first step was to collect three different datasets: Novels, News, and Twitter. Afterward, the Word2Vec algorithm is trained on many Persian datasets and used as word embedding. To have syntactic features besides semantic ones, Part Of Speech (POS) embedding has also been used. In the next step, an attention mechanism and Long Short-Term Memory (LSTM) are coupled with residual connections. Thus, the current layer input becomes the output of all previous layers rather than the top one. The model implemented is superior in performance for all three types of data compared to earlier models in text classification and author attribution. Absolute accuracies of 90.22%, 85.10%, and 71.39% were achieved among the three Novels, News, and Tweets datasets.

In the Internet of Things (IoT) era, Mobile Edge Computing (MEC) significantly enhances the efficiency of smart devices but is limited by battery life issues. Wireless Power Transfer (WPT) addresses this issue by providing a stable energy supply. However, effectively managing overall energy consumption remains a critical and under-addressed aspect for ensuring the network’s sustainable operation and growth. In this paper, we consider a WPT-MEC network wit user cooperation to migrate the double near-far effect for the mobile node (MD) far from the base station. We formulate the problem of maximizing long-term computation rates under a power consumption constraint as a multi-stage stochastic optimization (MSSO) problem. This approach is tailored for a sustainable WPT-MEC network, considering the dynamic and varying MEC network environment, including randomness in task arrivals and fluctuating channels. We introduce a virtual queue to transform the time-average energy constraint into a queue stability problem. Using the Lyapunov optimization technique, we decouple the stochastic optimization problem into a deterministic problem for each time slot, which can be further transformed into a convex problem and solved efficiently. Our proposed algorithm works efficiently online without requiring further system information. Rigorous mathematical analysis shows that our algorithm achieves O(1/V),O(V) trade-off between computation rate and queues stability. Extensive simulation results demonstrate that our proposed algorithm outperforms baseline schemes, achieving approximately 4% enhancement while maintain the queues stability.

This article introduces a new method for modelling and verifying the execution of timed security protocols (TSP) and their time-dependent security properties. The method, which is both novel and reliable, uses an extension of interpreted systems, accessible semantics in multi-agent systems, and timed interpreted systems (TIS) with dense time semantics to model TSP executions. We enhance the models of TSPs by incorporating delays and varying lifetimes to capture real-life aspects of protocol executions. To illustrate the method, we model a timed version of the Needham-Schroeder Public Key Authentication Protocol. We have also developed a new SMT-based bounded model checking reachability algorithm for the proposed structures and implemented it with the tool. The method comprises a new procedure for modelling TSP executions, a translation of TSP into TIS, and a translation of TIS's reachability problem into the SMT problem. The paper also includes thorough experimental results for nine protocols modelled by TIS and discusses the findings in detail.

Within environmental perception, automatic navigation and object detection, computer vision is a crucial and demanding field with many applications in modern industries, such as multi-target long-term visual tracking in automated production, defect detection, and driverless robotic vehicles with sensing ability. The performance of computer vision has greatly improved recently thanks to developments in deep learning algorithms and hardware computing capabilities, which has spawned the creation of a large number of related applications. This paper presents the results of a detailed review of over 50 papers published over the course of two decades (1999–2024), with a primary focus on the technical advancement of computer vision. To elucidate the foundational principles, an examination of typical visual algorithms based on traditional correlation filtering was initially conducted. Subsequently, a comprehensive overview of the most recent advancements in deep learning-based computer vision techniques was compiled. Furthermore, a comparative analysis between conventional and novel algorithms was undertaken to discuss the future trends and directions of computer vision. Lastly, the feasibility of employing visual SLAM (Simultaneous Localization and Mapping) algorithms in the context of autonomous vehicles with sensing capability was explored. Furthermore, we explored a thorough and efficient architecture for utilizing these techniques in autonomous robotic vehicles into unmanned green and low-carbon smart factories, underlining novel advances as well as potential prospects for future research.

Unmanned aerial vehicles (UAVs) have increasingly become integral to multi-access edge computing (MEC) due to their flexibility and cost-effectiveness, especially in the B5G and 6G eras. This paper aims to enhance the Quality of Experience (QoE) in large-scale UAV-MEC networks by minimizing the shrinkage ratio through optimal decision-making in computation mode selection for each user device (UD), UAV flight trajectory, bandwidth allocation, and computing resource allocation at edge servers. However, the interdependencies among UAV trajectory, binary task offloading mode, and computing/network resource allocation across numerous IoT nodes pose significant challenges. To address these challenges, we formulate the shrinkage ratio minimization problem as a mixed-integer nonlinear programming (MINLP) problem and propose a two-tier optimization strategy. To reduce the scale of the optimization problem, we first design a low-complexity UAV partition coverage algorithm based on the Welzl method and determine the UAV flight trajectory by solving a Traveling Salesman Problem (TSP). Subsequently, we develop a coordinate descent (CD) based method and an alternating direction method of multipliers (ADMM) based method for network bandwidth and computing resource allocation in the MEC system. The CD-based method is simple to implement and has a low computational complexity, while the ADMM-based method can further enhance the optimization result through joint optimization. Extensive simulations demonstrate that our proposed algorithms perform well in large-scale edge networks and outperform other representative benchmark methods.

This study systematically addresses the challenge of accurately modeling memristors, focusing on four distinct types doped with tungsten, tin, chromium, and carbon, fabricated by Known Inc. A comprehensive characterization was performed by subjecting the devices to sinusoidal excitations with varying frequencies and amplitudes, followed by data averaging and high-frequency filtering. The resulting measurements were fitted using three prominent memristor models: VTEAM, MMS, and Yakopcic, with additional bespoke modifications assessed. These models, typically formulated as coupled algebraic-differential equations integrating electrical quantities (voltage and current) with internal state variables governing device dynamics, were optimized using two robust approaches: (1) interior-point optimization with gradient-based search, and (2) Nelder-Mead gradient-free optimization, both with box constraints applied. A thorough comparison and discussion of the optimized model parameters ensued, accompanied by an examination of sensitivity to diverse frequency and amplitude ranges. The findings inform conclusions and provide a foundation for future refinements, underscoring the importance of multi-model evaluation and advanced optimization strategies in precise memristor modeling. The presented methodology offers a valuable framework for elucidating optimal modeling paradigms tailored to specific memristor architectures and operating regimes, ultimately enhancing their integration in emerging neuromorphic and computational applications.

Electromagnetic interference (EMI) generated by switching operations within high voltage DC (HVDC) converter stations is an issue that has been addressed by CIGRE and methods to manage EMI should be considered in the design phase using electromagnetic modeling techniques. It is shown that method of moments-based techniques that have been used in many previous studies may not be sufficient. Here, the finite-difference time-domain method is used to study the properties of single switching events using a realistic model of an HVDC converter station with special emphasis on determining the impact of the valve hall on shielding EMI. It is shown that the valve hall does not confine high frequency electromagnetic fields to the valve hall; rather it delays them from exiting through bushings in the wall. Further, it is shown that incorporating electromagnetic absorbing material into the valve hall design can significantly reduce EMI outside the converter station.

Loneliness is becoming prevalent across Japan in spite of its a dense population and communal culture. Surveys show 40% of adults report isolation across demographics. The contradiction emerges due to deteriorating social structure with modernization, economic parity, urbanization, and fading conventional setup. The continually social withdrawals, called “hikikomori." Relief seekers have embraced AI based emotive robots, digital avatars, and immersive VR as low-stakes social surrogates despite limited capacity to cultivate human intimacy. To stabilise government attempts awareness through reforms, scheme and support services. But a concrete transformation can be brough in with work-life balance, gender equality and rewire the metrics for self-esteem. Even a perfectly personalized AI cannot replace the human affection lacking in the society. Lasting solutions demand technological aids be balanced by confronting and reinventing community purpose and bring in structural change. Overall Japan’s endemic isolation reveals the societal fraying when modernization privileges independence over interdependence. The path forward rests on restoring cultural foundations to reverse loneliness and social isolation.

Malaria and Typhoid fever are prevalent diseases in tropical regions, exacerbated by unclear protocols, drug resistance, and environmental factors. Prompt and accurate diagnosis is crucial to improve accessibility and reduce mortality rates. Traditional diagnosis methods cannot effectively capture the complexities of these diseases due to the presence of confusable symptoms. Although machine learning (ML) models offer accurate predictions, they operate as "black boxes" with non-interpretable decision-making processes, making it challenging for healthcare providers to comprehend how the conclusions are reached. This study employs explainable AI (XAI) models such as Local Interpretable Model-agnostic Explanations (LIME), and Large Language Models (LLMs) like ChatGPT, Gemini, and Perplexity, to clarify diagnostic results for healthcare workers, building trust and transparency in medical diagnostics. The models were implemented on Google Colab and Visual Studio Code because of their rich libraries and extensions. Results showed that the Random Forest model outperformed Extreme Gradient Boost and Support Vector Machines and important features for predicting Malaria and Typhoid were identified with the LIME plots. Among LLMs, ChatGPT 3.5 demonstrates a comparative advantage over Gemini and Perplexity, hence highlighting how crucial it is to use a hybrid strategy that combines ML, XAI, and LLMs to improve diagnostic performance and reliability in healthcare applications. The study recommends implementing the integrated ML, LIME, and LLM processes because it streamlines the development and maintenance process's workflow overall, makes use of resources more effectively, and improves explainability by allowing the LLM to consider the complete context that LIME provides.

Multimodal emotion classification (MEC) involves analyzing and identifying human emotions by integrating data from multiple sources, such as audio, video, and text. This approach leverages the complementary strengths of each modality to enhance the accuracy and robustness of emotion recognition systems. However, one significant challenge is effectively integrating these diverse data sources, each with unique characteristics and levels of noise. Additionally, the scarcity of large, annotated multimodal datasets in Bangla limits the training and evaluation of models. In this work, we unveiled a pioneering multimodal Bangla dataset, MAViT-Bangla (Multimodal Audio Video Text Bangla dataset). This dataset, comprising 1002 samples across audio, video, and text modalities, is a unique resource for emotion recognition studies in the Bangla language. It features emotional categories such as anger, fear, joy, and sadness, providing a comprehensive platform for research. Additionally, we developed a framework for emotion recognition that employs a cross-modal attention mechanism among unimodal features. This mechanism fosters the interaction and fusion of features from different modalities, enhancing the model’s ability to capture nuanced emotional cues. The effectiveness of this approach was demonstrated by achieving an F1-score of 0.64, a significant improvement over unimodal methods.

Existing methods of hyperspectral anomaly detection still face several challenges: 1) Due to the limitations of self-supervision, avoiding the identity mapping of anomalies remains difficult; 2) The ineffective interaction between spatial and spectral features leads to the insufficient utilization of spatial information; 3) Current methods are not adaptable to the detection of multi-scale anomaly targets. To address the aforementioned challenges, we proposed a blind spot network based on multi-scale blind spot convolution for HAD. The multi-scale mask convolution module is employed to adapt to diverse scales of anomaly targets, while the dynamic fusion module is introduced to integrate the advantages of mask convolutions at different scales. The proposed approach includes a spatial-spectral joint module and a background feature attention mechanism to enhance the interaction between spatial-spectral features, with a specific emphasis on highlighting the significance of background features within the network. Furthermore, we propose a preprocessing technique that integrates pixel shuffle down-sampling (PD) operation with spatial spectral joint screening to address the issue of anomalous identity map-ping and facilitate finite-scale mask convolution to adapt to the detection of more scale targets. The proposed approach was assessed on four real hyperspectral datasets comprising anomaly targets of different scales. The experimental results demonstrate the effectiveness and superior performance of the proposed methodology compared with nine state-of-the art methods.

Over the past decades, overweight and obesity has become a global epidemic and the leading threat for death. To prevent the serious risk, an overweight or obese individual must apply a long-term weight-management strategy to control food intake and physical activities, which is however, not easy. Recently, with the advances of information technology, more and more people can use wearable devices and smartphones to obtain physical activity information, while they can also access various health-related information from Internet online social networks (OSNs). Nevertheless, there is a lack of an integrated approach that can combine these two methods in an efficient way. In this paper, we address this issue and propose a novel mobile-social framework for health recognition and recommendation, namely, H-Rec². The main ideas of H-Rec² include (1) to recognize the individual's health status using smartphone as a general platform, and (2) to recommend physical activity and food intake based on personal health information, life science principles, and health-related information obtained from OSNs. To demonstrate the potentials of the H-Rec² framework, we develop a prototype that consists of four important components: (1) an activity recognition module that senses physical activity using accelerometer, (2) a health status modeling module that applies a novel algorithm to generate personalized health status index, (3) a restaurant information collection module that collects relevant information from OSN, and (4) a restaurant recommendation module that provides personalized and context-aware recommendation. To evaluate the prototype, we conduct both objective and subjective experiments, which confirm the performance and effectiveness of the proposed system.

The paper addresses the energy consumption minimization problem in wireless powered communication networks (WPCNs) and proposes a DT-SWIPT-assisted time allocation scheme, named DaTA, such that the wireless station can share the remaining energy after transmission to the HAP to those who have not transmitted to the HAP to minimize the energy consumption of the network. In addition to proposing a new frame structure, the paper also considers the SNR (Signal-to-Noise) constraint to guarantee that the HAP can successfully receive data from wireless stations. In the paper, the energy consumption minimization problem is formulated as a nonlinear programming model. We employ the SQP (Sequential Quadratic Programming) algorithm to figure out the optimal solution. Moreover, a heuristic method is proposed as well to obtain a near-optimal solution in a shorter time. Simulation results showed that the proposed scheme can have better performance than the related work in terms of energy consumption and energy efficiency.

Proteomics research has become one of the most important topics in the fields of life science and natural science. At present, research on protein–protein interaction networks (PPINs) mainly focuses on detecting protein complexes or function modules. However, existing approaches are either ineffective or incomplete. In this paper, we investigate function module detection mechanisms in PPIN, including open databases, existing detection algorithms and recent solutions. After that, we describe the proposed solution based on simplified swarm optimization (SSO) algorithm and gene ontology knowledge. The proposed solution implements SSO algorithm for clustering proteins with similar function, and imports biological gene ontology knowledge for further identifying function complexes and improving detection accuracy. Furthermore, we use four different categories of species dataset for experiment: Fruitfly, Mouse, Scere, and Human. The testing and analysis result show that the proposed solution is feasible, efficient and could achieve a higher accuracy of prediction than existing approaches.

The rapid development of specific-purpose Large Language Models (LLMs), such as Med-PaLM, MEDITRON-70B, and Med-Gemini, has significantly impacted healthcare, offering unprecedented capabilities in clinical decision support, diagnostics, and personalized health monitoring. This paper reviews the advancements in medicine-specific LLMs, the integration of Retrieval-Augmented Generation (RAG) and prompt engineering, and their applications in improving diagnostic accuracy and educational utility. Despite the potential, these technologies present challenges, including bias, hallucinations, and the need for robust safety protocols. The paper also discusses the regulatory and ethical considerations necessary for integrating these models into mainstream healthcare. By examining current studies and developments, this paper aims to provide a comprehensive overview of the state of LLMs in medicine and highlight the future directions for research and application. The study concludes that while LLMs hold immense potential, their safe and effective integration into clinical practice requires rigorous testing, ongoing evaluation, and continuous collaboration among stakeholders.

Artificial Intelligence (AI) and Deep Reinforcement Learning (DRL) have revolutionised e-learning by creating personalised, adaptive, and secure environments. However, challenges such as privacy, bias, and data limitations persist. E-FedCloud aims to address these issues by providing more agile, personalised, and secure e-learning experiences. This study introduces E-FedCloud, an AI-assisted adaptive e-learning system that automates personalised recommendations and tracking, thereby enhancing student performance. It employs federated learning-based authentication to ensure secure and private access for both course instructors and students. Intelligent Software Agents (ISAs) evaluate weekly student engagement using the Shannon Entropy method, classifying students into either engaged or not-engaged clusters. E-FedCloud utilises weekly engagement status, demographic information, and an innovative DRL-based early warning system, specifically ID2QN, to predict the performance of not-engaged students. Based on these predictions, the system categorises students into three groups: risk of dropping out, risk of scoring lower in the final exam, and risk of failing the end exam. It employs a multi-disciplinary ontology graph and an attention-based capsule network for automated, personalised recommendations. The system also integrates performance tracking to enhance student engagement. Data is securely stored on a blockchain using the LWEA encryption method.

Brain is the most complex organ in the human body and it is divided into two hemispheres - left and right hemispheres. Left hemisphere is responsible for control of right side of our body whereas right hemisphere is responsible for control of left side of our body. Brain image segmentation from different neuroimaging modalities is one of the important parts in clinical diagnostic tools. Neuroimaging based digital imagery generally contain noise, inhomogeneity, aliasing artifacts, and orientational deviations. Therefore, accurate segmentation of brain images is a very difficult task. However, the development of accurate segmentation of brain images is very important and crucial for a correct diagnosis of any brain related diseases. One of the fundamental segmentation tasks is to identify and segment inter-hemispheric fissure/mid-sagittal plane, which separate the two hemispheres of the brain. Moreover, the symmetric/asymmetric analyses of left and right hemispheres of brain structures are important for radiologists to analyze diseases such as Alzheimer's, Autism, Schizophrenia, Lesions and Epilepsy. Therefore, in this paper we have analyzed the existing computational techniques used to find brain symmetric/asymmetric analysis in various neuroimaging techniques (MRI/CT/PET/SPECT), which are utilized for detecting various brain related disorders.

Abstract: With the growing integration of the Internet of Things (IoT) in smart buildings, it is crucial to ensure the precise implementation and operation of wireless sensor networks (WSNs). This paper aims to study the implementation aspect of Wireless Sensor Networks (WSNs) in a commercial multi-story building, specifically addressing the difficulty of dealing with variable environmental conditions on each floor. This research addresses the disparity between simulated situations and actual deployments, offering valuable insights that have the potential to signifi-cantly improve the efficiency and responsiveness of building management systems. We obtain real-time sensor data to analyze and evaluate the system’s performance. Our investigation is grounded in the growing importance of incorporating WSNs into buildings to create intelligent environments. By leveraging the capabilities of Contiki operating system, this study employs a specialized simulator to generate simulated data, allowing for a comparison between real-world and simulated scenarios. We provide an in-depth analysis to scrutinize the disparities and commonalities between the datasets obtained from real-world deployments and simulated en-vironments. We highlight the necessity of precise simulation to validate real-life sensor data, shedding light on the challenges and opportunities associated with implementing WSNs in in-telligent buildings. Results obtained show the significance of accurate simulation models for re-liable data representation, providing a roadmap for further developments in the integration of WSNs into intelligent building scenarios. The findings highlight the potential for optimizing living and working conditions based on the real-time monitoring of critical environmental pa-rameters. This includes insights into temperature, humidity, and light intensity, offering oppor-tunities for enhanced comfort and efficiency in intelligent environments.

This paper presents an analysis of the power flow within the Northern Interconnected Grid of Cameroon. The Newton-Raphson method has been performed, known for its accuracy, under MATLAB software, to model and solve complex power flow equations. This study simulates a series of outage scenarios to evaluate the responsive-ness of the grid. The results obtained underline the crucial importance of reactive power management and highlight the urgent need to consolidate the grid infrastructure of North Cameroon. To increase grid resilience and stability, the paper recommends the strategic integration of renewables and the development of interconnections with other power grids. These measures are presented as viable solutions to meet cur-rent and future energy distribution challenges, ensuring a reliable and sustainable power supply for Cameroon.

This paper proposed emotion recognition methods for consecutive facial images, and implements the inference of neural network model on a field-programmable gate array (FPGA). The proposed emotion recognition methods are based on a neural network model architecture that combines convolution neural networks (CNNs), long short-term memory (LSTM), and fully connected neural networks (FCNNs), called CLDNN or ConvLSTM-FCN. This type of neural network mod-el can analyze the local feature sequences obtained through convolution of data, making it suita-ble for processing time-series data such as consecutive facial images. In this paper, sequences of facial images are sampled from videos corresponding to various emotional state of subjects. The sampled images are then pre-processed with the processes includes facial detection, greyscale conversion, resize, and data augmentation if necessary. The 2-D CNN in ConvLSTM-FCN is used for feature extraction for these pre-processed facial images. These sequences of facial im-age features are time sequences with dependent properties between the elements within them. The LSTM is then used to model these time sequences followed by fully connected neural net-works (FCNNs) for classification. The proposed consecutive facial emotion recognition method achieves an average recognition rate of 99.51% on RAVDESS dataset, 87.80% on BAUM-1s dataset and 96.82% on eNTERFACE’05 data set, using 10-fold cross-validation on the PC. Some com-parisons of recognition accuracies between the proposed method and the other existing related works are conducted in this paper. According to the comparisons, the proposed emotion recog-nition methods outperform the existing related researches. The proposed emotion recognition methods are then implemented on an FPGA chip using the neural network model inference algo-rithms in this paper, and the accuracies of the experiments conducted on the FPGA chip are iden-tical to those obtained on the PC. This verifies that the proposed neural network model imple-mented on the FPGA chip performs well.

Few-shot object detection (FSOD) aims to address the challenge of requiring a substantial amount of annotations for training in conventional object detection, which is very labor-intensive. However, existing few-shot methods achieve high precision with the sacrifice of time-consuming for exhaustive fine-tuning, or take poor performance in novel-class adaptation. We presume the major reason is that the valuable correlation feature among different categories is insufficiently exploited, hindering the generalization of knowledge from base to novel categories for object detection. In this paper, we propose Few-Shot object detection via Correlation-RPN and Transformer Encoder-Decoder (CRTED), a novel training network to learn object-relevant features of inter-class correlation and intra-class compactness while suppressing object-agnostic features in the background with limited annotated samples. And we also introduce a 4-way tuple-contrast training strategy to positively activate the training progress of our object detector. Experiments over two few-shot benchmarks (Pascal VOC, MS-COCO) demonstrate that, our proposed CRTED without further fine-tuning can achieve comparable performance with current state-of-the-art fine-tuned works. The codes and pre-trained models will be released.

We present a very computationally light and fast approximation algorithm and then verify it with genetic algorithm and simulated annealing. We show that our algorithm is on par with GA and SA in terms of output produced while having a tightly bounded time complexity. Our algorithm works best when there is a strong positive correlation between the reliability of a component and its cost. We present two algorithms with the same essence. One of them is system cost bounded and the other is target reliability bounded. Our proposed algorithm works on a subsystem level redundancy instead of component level redundancy.

Deuterium is a heavy nonradioactive isotope of hydrogen, having a neutron as well as a proton, making it twice as heavy. It is a natural element, present at 156 parts per million in seawater. The ATPase pumps in the mitochondria utilize proton motive force to drive ATP synthesis, and deuterons damage the pumps, producing a stutter that can cause reactive oxygen release and inefficiencies in ATP synthesis. Cellular metabolism incorporates several novel mechanisms to assure low deuterium content in the mitochondria and other organelles. Nicotinamide adenine dinucleotide (NAD) is a major carrier of deuterium depleted (deupleted) protons to supply the mitochondria. Many enzymes, especially flavoproteins, are able to use proton tunneling to fractionate out deuterium. In this paper, we argue that the amino acid proline is able to trap and sequester deuterium, and that peptidyl prolyl isomerases (PPIases) play a central role in facilitating deuterium trapping in proline-rich proteins, most notably collagen. The endothelial glycocalyx also sequesters deuterium in gelled water lining the blood vessels, creating a battery and promoting low deuterium in the circulation. Excess deuterium promotes cancer growth, and cancer cells release large quantities of lactate via aerobic glycolysis to help reverse deuterium overload, systemically.

Mobile Edge Computing (MEC) integrated with Wireless Power Transfer (WPT) is emerging as a promising solution to reduce task delays and extend the battery life of Mobile Devices (MDs). Cooperative user communication, or relay technology, enhances communication efficiency for users located far from the base station by mitigating the double near-far effect caused by distance. This is considered a key technology in Beyond 5G (B5G) and future communication systems. In this paper, we focus on maximizing the long-term energy efficiency (EE) of a user-cooperation WPT-MEC system, while taking into account the uncertain load dynamics at the edge MD and the time-varying state of the wireless channel. The joint optimization of wireless charging time fraction, MD offloading duration, helper node processing time, and transfer power decision presents significant challenges due to the coupling of data offloading among cooperative users and a volatile system environment. To address these challenges, we formulate the problem as a stochastic programming problem and propose an online control algorithm, DOUCA, to solve it. Our approach utilizes Dinkelbach’s method and Lyapunov optimization theory to decouple the sequential decision problem into a deterministic sub-problem for each time slot. For the sub-problem, we use variable substitution to convert the non-convex problem into a convex one, containing only a small number of variables, which can be efficiently solved. Furthermore, we provide a mathematical analysis of our algorithm’s performance. Extensive simulation results demonstrate the effectiveness of our proposed algorithm, as evidenced by an impressive energy efficiency improvement of over 20% compared to benchmark methods. Our algorithm also achieves a trade-off between EE and system stability.

The modern supply chain (SC) is growing in terms of data, devices, users, and stakeholders, which introduced new security challenges and threats, especially with the reliance on centralized servers or cloud platforms. In addition, increased trust among system participants exposes the SC to a higher risk of vulnerabilities which require strong security measures. This article proposes a hybrid security framework for SC systems, BC-DRLzSC, that integrates Blockchain (BC) and Deep Reinforcement Learning (DRL) designed to operate in a zero trust (ZT) environment. In particular, we propose a decentralized BC-based approach integrated with smart contracts to manage system participant registration and authentication and to control access to system resources. BC-DRLzSC adopts a ZT architecture to reinforce SC security, which can be achieved with an advocate to verify each entity’s trustworthiness before granting or retaining access to system resources. Incorporating the ZT architecture, with BC and DRL, can potentially and significantly bolster SC system security. DRL is employed to develop a proactive attack detection model that continuously monitors the incoming traffic from authenticated nodes within the network and predicts any malicious actions. Finally, we evaluate the performance of our proposed DRL solution using the NSL-KDD dataset.

Remote Photoplethysmography (rPPG) has emerged as a non-intrusive and promising physiological sensing capability in HCI research, gradually extending its applications in health-monitoring and clinical care contexts. With advanced machine learning models, recent datasets collected in real-world conditions have gradually enhanced the performance of rPPG methods in recovering heart-rate and heart-rate variability metrics. However, the signal quality of reference ground-truth PPG data in existing datasets is by and large neglected, while poor quality references negatively influence models. Here, this work introduces a new imaging blood volume pulse (iBVP) dataset of synchronized RGB and thermal infrared videos with PPG ground-truth signals from the ear and its high resolution signal quality labels, for the first time. Participants perform rhythmic breathing, head-movement, and stress-inducing tasks, which help reflect real-world variations in psycho-physiological states. This work conducts dense (per sample) signal quality assessment to discard noisy segments of ground-truth and corresponding video frames. We further present a novel end-to-end machine learning framework, iBVPNet that features an efficient and effective spatio-temporal feature aggregation for reliable estimation of BVP signals. Finally, this work examines the feasibility of extracting BVP signals from thermal video frames, which is underexplored. The iBVP dataset and source codes are publicly available for research use.

Cloud computing refers to applications delivered as services over the Internet. Cloud systems employ policies that are inherently dynamic in nature and that depend on temporal conditions defined in terms of external events, such as the measurement of bandwidth, use of hosts, intrusion detection or specific time events. In this paper, we investigate an optimized resource management scheme named v-Mapper. The basic premise of v-Mapper is to exploit application-awareness concepts using software-defined networking (SDN) features. This paper makes three key contributions to the field: (1) We propose a virtual machine (VM) placement scheme that can effectively mitigate the VM placement issues for data-intensive applications; (2) We propose a validation scheme that will ensure that a service is entertained only if there are sufficient resources available for its execution and (3) We present a scheduling policy that aims to eliminate network load constraints. An evaluation was carried out with various benchmarks and demonstrated that v-Mapper shows improved performance over other state-of-the-art approaches in terms of average task completion time, service delay time and bandwidth utilization. Given the growing importance of supporting large-scale data processing and analysis in datacentres, the v-Mapper system has the potential to make a positive impact in improving datacentre performance in the future.

Mulberry leaves, when infected by pathogens, can suffer significant yield loss or even death if early disease detection and timely spraying are not performed. To enhance the detection performance of mulberry leaf diseases in natural environments and to precisely locate early small lesions, we propose a high-precision, high-efficiency disease detection algorithm named YOLOv8-RFMD. Based on improvements to You Only Look Once version 8 (YOLOv8), we first proposed the Multi Dimension Feature Attention (MDFA) module, which integrates important features at the pixel-level, spatial and channel dimensions. Building on this, we designed the RFMD Module, which consists of the Conv-BatchNomalization-SiLU (CBS) module, Receptive-Field Coordinated Attention (RFCA) Conv, and MDFA, replacing the Bottleneck in the model’s Residual block. We then employed Adown down sampling structure to reduce the model size and computational complexity. Finally, to improve the detection precision of small lesion features, we replaced the complete intersection over union (CIOU) loss function with the Normalized Wasserstein Distance (NWD) loss function. Results show that the YOLOv8-RFMD model achieves an mAP50 of 94.3% and an mAP50:95 of 67.8% on experimental data, representing increases of 2.9% and 4.3% respectively compared to the original model. The model size is reduced by 0.53 MB to just 5.45 MB, and the number of floating-point operations is reduced by 0.3 G to only 7.8 G. The improved model meets the deployment requirements of mobile embedded devices and can provide a theoretical reference for the automated spraying operations of mulberry leaves.

In emergency situations, ensuring standardized Cardiopulmonary Resuscitation (CPR) actions is crucial. However, current Automated External Defibrillators (AEDs) lack methods to determine whether CPR actions are performed correctly, leading to inconsistent CPR quality. To address this issue, we introduce a novel method called Deep Learning-based CPR Action Standardization (DLCAS). This method involves three parts. First, it detects correct posture using OpenPose to recognize skeletal points. Second, it identifies a marker wristband with our CPR-Detection algorithm and measures compression depth, count and frequency using a depth algorithm. Finally, we optimize the algorithm for edge devices to enhance real-time processing speed. Extensive experiments on our custom dataset have shown that the CPR-Detection algorithm achieves a mAP0.5 of 97.04%, while reducing parameters to 0.20M and FLOPs to 132.15K. In a complete CPR operation procedure, the depth measurement solution achieves an accuracy of 90% with a margin of error less than 1 cm, while the count and frequency measurements achieve 98% accuracy with a margin of error less than 2 counts. Our method meets the real-time requirements in medical scenarios, and the processing speed on edge devices have increased from 8fps to 25fps.

Globalization and industrialization have significantly disturbed the environmental ecosystem, leading to critical challenges such as global warming, extreme weather events, and water scarcity. Forecasting temperature trends is crucial for enhancing the resilience and quality of life in smart sustainable cities, enabling informed decision-making and proactive urban planning. This research specifically targets Ahmedabad city in India and employs the Seasonal Auto Regressive Integrated Moving Average with eXogenous factors (SARIMAX) model to forecast temperatures over a ten-year horizon using two decades of real-time temperature data. The stationarity of the dataset was confirmed using the Augmented Dickey-Fuller test, and the Akaike Information Criterion (AIC) method helped identify the optimal seasonal parameters of the model, ensuring a balance between fidelity and predictive accuracy. The model achieved an RMSE of 1.0265, indicating high accuracy within the typical range for urban temperature forecasting. This robust measure of error underscores the model’s precision in predicting temperature deviations, which is particularly relevant for urban planning and environmental management. The findings provide city planners and policymakers with valuable insights and tools for preempting adverse environmental impacts, marking a significant step towards operational efficiency and enhanced governance in future smart urban ecosystems. Future work may extend the model's applicability to broader geographical areas and incorporate additional environmental variables to refine predictive accuracy further.

This paper introduces a novel decision-making approach grounded in insights into human visual perception of change. Modern technologies such as internet of things (IoT) provide us with large amounts of sensor data that need to be processed in real time and decisions made with a high degree of accuracy and reliability. Artificial intelligence (AI) methods are welcome in this context and need to be upgraded to meet actual challenges. While modern computing capabilities facilitate rapid data processing, the real-time demands of vast sensor data necessitate swift responses across the cyber chain, often leading to compromises in solution quality to circumvent combinatorial search complexities. Determining the adequacy of a solution entails varied approaches, often relying on heuristic methodologies. We illustrate our original approach with an example of a selected detail of a differential evolution algorithm, where we have to make a decision to adopt the best solution so far. We propose an approach inspired by human perceptual features that exploits the Weber-Fechner law to emulate human judgements, and offers a promising way to improve decision making in AI applications and real-time requirements fulfillment. Our proposed methodology demonstrates applicability across diverse AI scenarios involving numerical data, effectively mirroring human perception abilities.

This paper compares Demand Response algorithms suitable for Short-term Forecasting based on Advanced Metering Infrastructure (AMI). It selects ARIMA (AutoRegressive Integrated Moving Average), SARIMA (Seasonal ARIMA), LSTM (Long Short-Term Memory), and SVM (Support Vector Machine) as the primary research algorithms and delves into their theoretical foundations and functional characteristics. The theoretical background section explores each algorithm’s applicability in analyzing time-series data, recognizing nonlinear patterns, and solving optimization and classification problems. The methodology section will detail the methods of AMI data collection, the implementation of the algorithms, and the experimental setup. In the experiments and results section, the performance of each algorithm will be assessed by applying AMI data and presenting the outcomes of comparative analyses, with a focus on evaluating the predictive accuracy of the algorithms. The discussion will interpret the experimental results, analyzing the advantages and disadvantages of each algorithm and their applicability in the AMI system. It will also explore the limitations of the current research and directions for future research. The conclusion will summarize the key findings of the study and affirm how this research contributes to understanding and improving short-term forecasting and demand response systems based on AMI.

The Internet of Things (IoT) is a concept based on the integration of various devices and methods of communication to exchange information in order to contribute to the improvement of everyday life, business and environment. This paper focuses on the establishment of a modern Internet of Things platform to ensure secure collection, processing and storage of data, under unexpected events. This platform establishes reliable connectivity which enhance system uptime; the key to success when implementing an IoT strategy. For this purpose, we proposed a hybrid approach combining different connectivity technologies to ensure the availability and smooth operation of the system and the prevention of data loss in unforeseen cases.

In this paper, a new approach to retrieve semantic images based on shape and geometric features of image in conjunction with multi-class support vector machine is proposed. Zernike moment as shape feature is to verify the invariance of objects for silhouette image. In addition, a set of geometrical features is to explore the objects shape using two features of rectangularity and circularity. Then the extracted features are normalized and employed for multi-class support vector machine either for learning or retrieving processes. The retrieving process relies on three main tasks which namely Query Engine, Matching Module and Ontology Manger, respectively. Query Engine is to build the input text or image query using SPARQL language. The matching module extracts the shape and geometric features of image’s objects and employ them to Ontology Manger which in turn inserts them in ontology knowledge base. Benchmark mammals have been conducted to empirically conclude the outcome of proposed approach. Our experiment on text and image retrieval yields efficient results to problematic phenomena than previously reported.

Household waste is the primary source of environmental pollution due to global population growth compared to other sources of waste. This research aims to develop a smart and integrated household waste management system framework. The resulting framework not only focuses on dimensions of information technology but also links it with other integrated dimensions. This research attempts to conduct a Systematic Literature Review (SLR) using the Preferred Reporting Items for Systematic Review and Meta-analysis (PRISMA) protocol to produce a smart and integrated household waste management framework. The framework's design is carried out by identifying the type of household waste management process based on the Integrated Sustainable Waste Management (ISWM) framework, dimensions that support smart household waste management, and stakeholders involved. The SLR results, namely dimensions and subdimensions that support the framework of a smart and integrated household waste management system, were validated by a resource person from the Indonesian Ministry of Environment and Forestry. The smart and integrated household waste management framework that has been developed in this research contains five main dimensions: Information Technology, Operational Infrastructure, Governance, Economy, and Social-culture. This framework also addresses stakeholder engagement to support smart household waste management systems and identifies waste management processes based on the ISWM framework. This research uses the PRISMA technique to provide a framework for smart and integrated household waste management systems in the initial stage. The proposed framework has been validated and can be further developed as a smart and integrated household waste management system. This research also contributes to the involvement of various dimensions identified to deal with waste problems.

Underwater imaging presents unique challenges, notably color distortions and reduced contrast due to light attenuation and scattering. Most underwater image enhancement methods, first use linear transformations for color compensation and then enhance the image. We observed that linear transformation for color compensation is not suitable for certain images. For such images, a non-linear mapping is a better choice. Our paper introduces a unique underwater image restoration approach leveraging a streamlined convolutional neural network (CNN) for dynamic weight learning for linear and non-linear mapping. In the first phase, a classifier is applied that classifies the input images into Type-I or Type-II. In the second phase, depending on the classification, the Deep Line Model (DLM) for Type-I images or the Deep Curve Model (DCM) for Type-II images. For mapping an input image to an output image, DML creatively combines color compensation and contrast adjustment in a single step and uses deep lines for transformation, whereas DCM employs higher-order curves. Both models utilize lightweight neural networks that learn per-pixel dynamic weights based on the input image’s characteristics. Comprehensive evaluations on benchmark datasets using metrics like peak signal-to-noise ratio(PSNR) and root mean squares error (RMSE) affirm our method’s effectiveness in accurately restoring underwater images, outperforming existing techniques.

In the fast-changing world of software development, the protection of software products has emerged as an imperative requirement. This paper presents a new way to increase the maturity of development teams to reach the highest level of software security. Based on these, the framework uses the Fuzzy Analytic Hierarchy Process (Fuzzy-AHP) to systematically evaluate and enhance the people maturity of software development projects. The essence of the fuzzy logic and AHP technique interaction is to handle the uncertainty and complexity of human factors and team dynamics in the evaluation framework. Using the decision-making model allows the project managers and stakeholders to determine the appropriate areas for improvement and develop the right strategies and actions to nurture a secure and mature development culture. The paper identifies 24 human success factors (HSFs) and human security vulnerabilities (HSVs) and 38 practices for addressing these HSFs and HSVs. Furthermore, we discuss the local and global ranks of each HSF and HSV practice and categorize the identified practices into 9 categories to determine the ranks and weight of each category. Based on collected data, FAHP prioritized these practices; the category “C4: Skill Development and Stakeholder Engagement” is ranked highest at rank-1 and possesses the most significant weight of 0.12435. Similarly, the highest global weight is 0.051506, and the global ranked (rank-1) HSF and HSV practice is “P15: Hands-On Practice and Stakeholder Communication”. Research evidence and case discussions show how the described framework assists in building secure software development (SSD) practices, which can be considered evidence of the application of team maturity to improve cybersecurity in organizations. Additional research directions include improving the framework, especially using highly developed learning techniques and applying the framework to other forms of development.

In this work, we introduce the first pipeline that combines a refraction-aware Structure-from-Motion (SfM) method with a deep learning model specifically designed for airborne bathymetry. We accurately estimate the 3D positions of the submerged points by integrating refraction geometry within the SfM optimization problem. This way, no refraction correction as post-processing is required. Experiments with simulated data that approach the real-world capturing conditions demonstrate that SfM with refraction correction is extremely accurate, with submillimeter errors. We integrate our refraction-aware SfM within a deep learning framework that also takes into account radiometrical information, developing a combined spectral and geometry-based approach, with further improvements on accuracy and robustness to different seafloor types, both textured and textureless. We conducted experiments with real-world data at 2 locations in the southern Mediterranean Sea, with varying seafloor types which demonstrate the benefits of refraction correction for the deep learning framework. We open-source our refraction-aware SfM, providing researchers in airborne bathymetry with a practical tool to apply SfM in shallow water areas.

Some of the barriers preventing Virtual Reality (VR) from being widely adopted are the cost and unfamiliarity of VR systems. Here we propose that in many cases the specialized controllers shipped with most VR head-mounted displays can be replaced by a regular smartphone, cutting the cost of the system, and allowing users to interact in VR using a device they are already familiar with. To achieve this, we developed SmartVR Pointer, an approach that uses smartphones as a replacement for the specialized controllers for two essential operations in VR: selection and navigation by teleporting. In SmartVR Pointer a camera mounted on the head-mounted display (HMD) is tilted downwards so that it points to where the user will naturally be holding their phone in front of them. SmartVR Pointer supports three selection modalities: tracker-based,gaze-based, and combined/hybrid. In the tracker-based SmartVR pointer selection we use image-based tracking to track a QR code displayed on the phone screen and them map the phone’s position to a pointer shown within the field of view of the camera in the Virtual Environment. In the gaze-based selection modality the user places the pointer using their gaze and taps on the phone for selection. The combined technique is a hybrid between head-gaze-based interaction in VR and smartphone-based Augmented Reality. It allows the user to control a VR pointer that behaves like a mouse pointer by moving their smartphone to select objects within the virtual environment, and to interact with the selected objects using the smartphone’s touch screen. The touchscreen is used for selection and dragging. SmartVR Pointer is simple and requires no calibration and no complex hardware assembly or disassembly. We demonstrate successful interactive applications of SmartVR Pointer in a VR environment with a demo where the user navigates in the virtual environment using teleportation points on the floor and then solves a Tetris-style key-and-lock challenge.

We introduce NodeFlow, a flexible framework for probabilistic regression on tabular data that combines Neural Oblivious Decision Ensemble (NODE) and Conditional Continuous Normalizing Flows (CNF). It offers improved modeling capabilities for arbitrary probabilistic distributions, addressing the limitations of traditional parametric approaches. In NodeFlow, the NODE captures complex relationships in tabular data through a tree-like structure, while the conditional CNF utilizes the NODE’s output space as a conditioning factor. The training process of NodeFlow employs standard gradient-based learning, facilitating end-to-end optimization of the NODEs and CNF-based density estimation. This approach ensures outstanding performance, ease of implementation, and scalability, making NodeFlow an appealing choice for practitioners and researchers. Comprehensive assessments on benchmark datasets underscore NodeFlow’s efficacy, revealing its achievement of state-of-the-art outcomes in multivariate probabilistic regression setup and its strong performance in univariate regression tasks. Furthermore, ablation studies are conducted to justify the design choices of NodeFlow. In conclusion, NodeFlow’s end-to-end training process and strong performance make it a compelling solution for practitioners and researchers. Additionally, it opens new avenues for research and application in the field of probabilistic regression on tabular data.

The advancements in social networking have empowered open expression on micro blogging platforms like Twitter. Traditional Twitter Sentiment Analysis (TSA) faced challenges due to rule-based or dictionary algorithms, dealing with feature selection, ambiguity, sparse data, and language variations. This study proposed a classification framework for Twitter sentiment data using word count vectorization and machine learning techniques to reduce the difficulties addressed on annotated sentiment-labelled tweets. Various classifiers (Naive Bayes, Decision Tree, K Nearest Neighbours, Logistic Regression, and Random Forest) were evaluated based on Accuracy, Precision, Recall, F1-score, and Specificity. Random Forest outperformed others with an Area under Curve (AUC) value of 0.96, and an Average Precision (AP) score of 0.96 in sentiment classification, especially effective with minimal Twitter-specific features.

Real-time object detection remains a pivotal topic in the realm of computer vision. Balancing the accuracy and speed of object detectors poses a formidable challenge for both academic researchers and industry practitioners. While recent transformer-based models have showncase the prowess of the attention mechanism, delivering remarkable performance improvements over CNNs, their computational demands can hinder their effectiveness in real-time detection tasks. In this paper, we elect to use YOLOX as our robust starting point and introduce a series of effective enhancements, culminating in a new high-performance detector named YOLOAX. To further exploit the power of the attention mechanism, we devise multi-dimensional attention-based modules that activate CNNs, emphasizing the most pertinent regions and bolstering the capacity to learn the most informative image representations from feature maps. Moreover, we introduce a novel leading label assignment strategy called STA, along with a groundbreaking loss function named GEIOU Loss, to further refine our detector's performance. We provide extensive ablation studies on the COCO and PASCAL VOC 2012 detection datasets to validate our proposed methods. Remarkably, our YOLOAX is trained solely on the MS-COCO dataset from scratch, without leveraging any prior knowledge, which achieves an impressive 54.2% AP on the COCO 2017 test set while maintaining a real-time speed of 72.3 fps, surpassing YOLOX by a significant margin of 3.0% AP. Our source code and pre-trained models are openly accessible at https://github.com/KejianXu/yoloax.

The purpose of this paper is to propose a novel transfer learning regularization method based on knowledge distillation. Recently, transfer learning methods have been used in various fields. However, the problems such as knowledge loss still occur during the process of transfer learning to a new target dataset. To solve these problems, there are various regularization methods based on knowledge distillation techniques. In this paper, we propose a transfer learning regularization method based on feature map alignment used in the field of knowledge distillation. The proposed method is composed of two attention based submodules: self-pixel attention (SPA) and global channel attention (GCA). The self-pixel attention submodule utilizes both the feature maps of the source and target models, so that it provides an opportunity to jointly consider the features of the target and the knowledge of the source. The global channel attention submodule determines the importance of channels through all layers, unlike the existing methods that calculate these only within a single layer. Accordingly, transfer learning regularization is performed by considering both the internal of each single layer and the depth of the entire layer. Consequently, the proposed method using both of these submodules showed overall improved classification accuracy than the existing methods in classification experiments on commonly used dataset.