IEEE Open Journal of the Computer Society最新文献

{"title":"Removing Neurons From Deep Neural Networks Trained With Tabular Data","authors":"Antti Klemetti;Mikko Raatikainen;Juhani Kivimäki;Lalli Myllyaho;Jukka K. Nurminen","doi":"10.1109/OJCS.2024.3467182","DOIUrl":"https://doi.org/10.1109/OJCS.2024.3467182","url":null,"abstract":"Deep neural networks bear substantial cloud computational loads and often surpass client devices' capabilities. Research has concentrated on reducing the inference burden of convolutional neural networks processing images. Unstructured pruning, which leads to sparse matrices requiring specialized hardware, has been extensively studied. However, neural networks trained with tabular data and structured pruning, which produces dense matrices handled by standard hardware, are less explored. We compare two approaches: 1) Removing neurons followed by training from scratch, and 2) Structured pruning followed by fine-tuning through additional training over a limited number of epochs. We evaluate these approaches using three models of varying sizes (1.5, 9.2, and 118.7 million parameters) from Kaggle-winning neural networks trained with tabular data. Approach 1 consistently outperformed Approach 2 in predictive performance. The models from Approach 1 had 52%, 8%, and 12% fewer parameters than the original models, with latency reductions of 18%, 5%, and 5%, respectively. Approach 2 required at least one epoch of fine-tuning for recovering predictive performance, with further fine-tuning offering diminishing returns. Approach 1 yields lighter models for retraining in the presence of concept drift and avoids shifting computational load from inference to training, which is inherent in Approach 2. However, Approach 2 can be used to pinpoint the layers that have the least impact on the model's predictive performance when neurons are removed. We found that the feed-forward component of the transformer architecture used in large language models is a promising target for neuron removal.","PeriodicalId":13205,"journal":{"name":"IEEE Open Journal of the Computer Society","volume":"5 ","pages":"542-552"},"PeriodicalIF":0.0,"publicationDate":"2024-09-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10693557","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142397170","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"AE-YOLOv5 for Detection of Power Line Insulator Defects","authors":"Wei Shen;Ming Fang;Yuxia Wang;Jiafeng Xiao;Huangqun Chen;Weifeng Zhang;Xi Li","doi":"10.1109/OJCS.2024.3465430","DOIUrl":"https://doi.org/10.1109/OJCS.2024.3465430","url":null,"abstract":"The power transmission network, which delivers power energy from generator to customers, plays an important role in the power grid. Insulator is a basic component in the power transmission network. Its defects may lead to the paralysis of the entire transmission network, resulting in serious electricity accidents. Therefore, how to use artificial intelligence and other emerging technologies to realize automatic detection of power line insulator defects has become an urgent problem to be solved. To accurately detect insulator defects in complex environment, this article proposes Attention Enhanced YOLOv5 (AE-YOLOv5) by inserting visual attention modules into original YOLOv5 model. In particular, we design a Channel-Spatial Attention module and plug it into the backbone of YOLOv5 to enhance its representation learning ability. Furthermore, a Multi-scale Attention module is also proposed to enhance the Feature Pyramid Network (FPN). To validate the efficacy of our proposed model, we conducted training and testing on a dataset collected from real-world scenarios. The experimental results demonstrate that our model can effectively and accurately detect defects of power line insulators in real-time.","PeriodicalId":13205,"journal":{"name":"IEEE Open Journal of the Computer Society","volume":"5 ","pages":"468-479"},"PeriodicalIF":0.0,"publicationDate":"2024-09-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10684881","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142368634","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Human Learning Optimization Algorithm With Diversified Searches","authors":"Jiamin Kou;Ke Li;Leyu Zheng","doi":"10.1109/OJCS.2024.3465444","DOIUrl":"https://doi.org/10.1109/OJCS.2024.3465444","url":null,"abstract":"A Human Learning Optimization with Diversified Search (DSHLO) algorithm is proposed to address the limitation of existing human learning optimization algorithms, such as smaller search space, and local optima due to the replication of optima in both individual and social learning operations. By introducing diversified search strategies, the DSHLO algorithm uses different methods to explore different solution spaces by simulating different human learning styles. Firstly, chaotic mapping is employed to enhance the population's likelihood of evolution. Secondly, inductive learning operators enrich the population diversity by combining learned individual and social knowledge with new one. Thirdly, the stochastic learning operator, based on the triangular walking strategy, increases the local optimization capability of the algorithm. Finally, the social learning operator, based on social hierarchy dominance, improves the convergence rate. The proposed algorithm is validated on the CEC2017 test set by comparison with nine baseline algorithms. The experimental results show that the DSHLO algorithm achieves faster convergence speeds and higher accuracy in most of the cases. Experiments on a supply chain planning and scheduling application prove that the proposed algorithm is also eligible to solve the practical engineering problems.","PeriodicalId":13205,"journal":{"name":"IEEE Open Journal of the Computer Society","volume":"5 ","pages":"589-598"},"PeriodicalIF":0.0,"publicationDate":"2024-09-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10684996","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142447171","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Exploring the Potential of Deep-Learning and Machine-Learning in Dual-Band Antenna Design","authors":"Rida Gadhafi;Abigail Copiaco;Yassine Himeur;Kiyan Afsari;Husameldin Mukhtar;Khalida Ghanem;Wathiq Mansoor","doi":"10.1109/OJCS.2024.3463190","DOIUrl":"10.1109/OJCS.2024.3463190","url":null,"abstract":"This article presents an in-depth exploration of machine learning (ML) and deep learning (DL) for the optimization and design of dual-band antennas in Internet of Things (IoT) applications. Dual-band antennas, which are essential for the functionality of current and forthcoming flexible wireless communication systems, face increasing complexity and design challenges as demands and requirements for IoT-connected devices become more challenging. The study demonstrates how artificial intelligence (AI) can streamline the antenna design process, enabling customization for specific frequency ranges or performance characteristics without exhaustive manual tuning. By utilizing ML and DL tools, this research not only enhances the efficiency of the design process but also achieves optimal antenna performance with significant time savings. The integration of AI in antenna design marks a notable advancement over traditional methods, offering a systematic approach to achieving dual-band functionality tailored to modern communication needs. We approached the antenna design as a regression problem, using the reflection coefficient, operating frequency, bandwidth, and voltage standing wave ratio as input parameters. The ML and DL models then are used to predict the corresponding design parameters for the antenna by using 1,000 samples, from which 700 are allocated for training and 300 for testing. This effectiveness of this approach is demonstrated through the successful application of various ML techniques, including Fine Gaussian Support Vector Machines (SVM), as well as Regressor and Residual Neural Networks (ResNet) with different activation functions, to optimize the design of a dual-band T-shaped monopole antenna, thereby substantiating AI's transformative potential in antenna design.","PeriodicalId":13205,"journal":{"name":"IEEE Open Journal of the Computer Society","volume":"5 ","pages":"566-577"},"PeriodicalIF":0.0,"publicationDate":"2024-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10683980","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142254393","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Artificial Intelligence of Behavior for Human Emotion Recognition in Closed Environments","authors":"Gonzalo-Alberto Alvarez-Garcia;Claudia Zúñiga-Cañón;Antonio-Javier Garcia-Sanchez;Joan Garcia-Haro;Milton Sarria-Paja;Rafael Asorey-Cacheda","doi":"10.1109/OJCS.2024.3463173","DOIUrl":"10.1109/OJCS.2024.3463173","url":null,"abstract":"Understanding human emotions and behavior in closed environments is essential for creating more empathetic and humane spaces. Environmental factors, such as temperature, noise, and light, play a crucial role in influencing behavior, but individuals' emotional states are equally important and often go unnoticed. Artificial Intelligence of Behavior (AIoB) offers a novel approach that integrates environmental measurements with human emotions to create spatially adaptive processes that can influence behavior. In this article, we present a new human emotion sensor developed using video cameras and implemented on a System on Chip (SoC) development board. Our approach uses Convolutional Neural Networks (CNNs) to recognize the presence of emotions in enclosed spaces and generate parameters that can influence emotional states and behavior within an AIoB system. The research successfully integrates advanced CNN technology into a System on Chip (SoC) platform, allowing for real-time processing of video data. The versatility of utilizing an energy-efficient SoC extends its application to smart environments aimed at improving mental health. By employing algorithms capable of detecting emotional states across various individuals, the study enhances its effectiveness. Additionally, it identifies the best CNN operations tailored to the technical specifications of the devices involved. Thus, The development involves a three-step process: (i) collecting enough data to build a robust model, (ii) training the model and evaluating its performance using test values, and (iii) applying the model on the development board. Our study demonstrates the feasibility of using AIoB to recognize and respond to human emotions in closed areas. By integrating emotional cues with environmental measurements, our system can create more personalized and empathetic spaces that cater to the needs of individuals. Our approach could have significant implications for designing public spaces to promote well-being and emotional satisfaction.","PeriodicalId":13205,"journal":{"name":"IEEE Open Journal of the Computer Society","volume":"5 ","pages":"578-588"},"PeriodicalIF":0.0,"publicationDate":"2024-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10683879","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142254389","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Medicine's New Rhythm: Harnessing Acoustic Sensing via the Internet of Audio Things for Healthcare","authors":"Farrukh Pervez;Moazzam Shoukat;Varsha Suresh;Muhammad Umar Bin Farooq;Moid Sandhu;Adnan Qayyum;Muhammad Usama;Anna Girardi;Siddique Latif;Junaid Qadir","doi":"10.1109/OJCS.2024.3462812","DOIUrl":"10.1109/OJCS.2024.3462812","url":null,"abstract":"In the modern landscape of information and communication technologies, the current healthcare industry confronts significant challenges. These include a shortage of experienced medical professionals, disparities in access to healthcare services that persist across different regions around the globe, and an increased need for detailed, real-time monitoring of patients in both urban and remote regions. This article delves into the potential of the Internet of Audio Things for Healthcare (IoAuT4H), which lies at the intersection of Internet of Audio Things and the Internet of Medical Things, as a solution to these pressing issues. By seamlessly merging cutting-edge audio technology, networking, and the advanced deep learning techniques, the IoAuT4H emerges as a promising solution. It has the potential to reshape hospital and clinical infrastructure, streamline early medical interventions, and facilitate rapid emergency responses. Additionally, this study underscores the pivotal role of the IoAuT4H in strengthening overall health practices, refining care methods for the elderly, and rejuvenating paediatric health approaches. While the benefits of the IoAuT4H are numerous, this article also critically examines the challenges in its widespread adoption. These include ethical considerations, ensuring the accuracy of audio data, and integrating it effectively with existing healthcare systems. In conclusion, this article seeks to provide a comprehensive understanding of the IoAuT4H, positioning it as a bridge between current healthcare challenges and technological advancements.","PeriodicalId":13205,"journal":{"name":"IEEE Open Journal of the Computer Society","volume":"5 ","pages":"491-510"},"PeriodicalIF":0.0,"publicationDate":"2024-09-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10683979","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142254390","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A Taxonomy-Based Survey of EM-SCA and Implications for Multi-Robot Systems","authors":"Yomna Mokhtar Ibrahim;Shabnam Kasra Kermanshahi;Kathryn Kasmarik;Jiankun Hu","doi":"10.1109/OJCS.2024.3461808","DOIUrl":"10.1109/OJCS.2024.3461808","url":null,"abstract":"Electromagnetic Side Channel Analysis (EM-SCA) is a major area of interest within the field of cybersecurity. EM-SCA makes use of the electromagnetic radiation that naturally leaks from any device that runs on electricity. Information about the observed device can be gained by gathering and analysing these electromagnetic traces. Numerous studies have demonstrated the applicability of this side channel in various environments for legal and illegal objectives. On the other hand, multi-robot systems, including swarm robotics, have received considerable attention in recent years due to their ability to conduct complex tasks using simple robots cooperating with each other. Although multi-robot and swarm robot systems are likely to be widely used in practical applications in the near future, security concerns in this context have not yet received enough attention. In particular, to the best of our knowledge, EM-SCA threats and benefits have never been thoroughly examined in this context before. In order to spotlight this matter, this work begins with a thorough introduction to EM-SCA and provides a taxonomic structure. Then, guided by this taxonomy, we present a range of EM-SCA scenarios that need to be considered in multi-robot applications.","PeriodicalId":13205,"journal":{"name":"IEEE Open Journal of the Computer Society","volume":"5 ","pages":"511-529"},"PeriodicalIF":0.0,"publicationDate":"2024-09-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10681247","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142254391","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Utilizing Deep improved ResNet50 for Brain Tumor Classification Based MRI","authors":"Karrar Neamah;Farhan Mohamed;Safa Riyadh Waheed;Waleed Hadi Madhloom Kurdi;Adil Yaseen Taha;Karrar Abdulameer Kadhim","doi":"10.1109/OJCS.2024.3453924","DOIUrl":"10.1109/OJCS.2024.3453924","url":null,"abstract":"A robust approach for brain tumor classification is being developed using deep convolutional neural networks (CNNs). This study leverages an open-source dataset derived from the MRI Brats2015 brain tumor dataset. Preprocessing included intensity normalization, contrast enhancement, and downsizing. Data augmentation techniques were also applied, encompassing rotations and flipping. The core of our proposed approach lies in the utilization of a modified ResNet-50 architecture for feature extraction. This model integrates transfer learning by replacing the final layer with a spatial pyramid pooling layer, enabling it to leverage pre-trained parameters from ImageNet. Transfer learning from ImageNet aids in countering overfitting. Our model's performance was evaluated with various hyperparameters, including existing methods in terms of accuracy, precision, recall, F1-score, sensitivity, and specificity. This study showcases the potential of deep learning, transfer learning, and spatial pyramid pooling in MRI-based brain tumor classification, providing an effective tool for medical image analysis. Our methodology employs a modified ResNet-50 architecture with transfer learning, integrating a spatial pyramid pooling layer for feature extraction. Systematic evaluation showcases the model's superiority over existing methods, demonstrating remarkable results in accuracy (0.9902), precision (0.9837), recall (0.9915), F1-score (0.9891), sensitivity, and specificity. The comparative analysis against prominent CNN architectures reaffirms its outstanding performance. Our model not only mitigates overfitting challenges but also offers a promising tool for medical image analysis, underlining the combined efficacy of spatial pyramid pooling and transfer learning. The study's optimization parameters, including 25 epochs, a learning rate of 1e-4, and a balanced batch size, contribute to its robustness and real-world applicability, furthering advancements in efficient brain tumor classification within MRI data.","PeriodicalId":13205,"journal":{"name":"IEEE Open Journal of the Computer Society","volume":"5 ","pages":"446-456"},"PeriodicalIF":0.0,"publicationDate":"2024-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10670206","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142208847","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Grasshopper Optimization Algorithm and Neural Network Classifier for Detection and Classification of Barley Leaf Diseases","authors":"O. Dorgham;G. Abu-Shareah;O. Alzubi;J. Al Shaqsi;S. Aburass;M. A. Al-Betar","doi":"10.1109/OJCS.2024.3457160","DOIUrl":"10.1109/OJCS.2024.3457160","url":null,"abstract":"The prevalence of plant diseases presents a substantial challenge to global agriculture, significantly impacting both production levels and economic stability in numerous countries. This study focuses on the early detection of two prevalent diseases affecting barley leaves: net blotch and spot blotch. We introduce a novel model designed for the accurate detection and classification of these diseases. The model employs advanced pre-processing techniques, including the transformation of images into the CIELAB color space and the segmentation of affected areas, to enhance disease identification accuracy. Key shape properties characterizing the diseased regions are extracted and analyzed to distinguish between the two diseases. A critical component of our approach is the feature selection phase, aimed at identifying the most pertinent and informative features, thereby minimizing classification errors and maximizing model accuracy with a minimal set of shape properties. To optimize this process, we have incorporated the Grasshopper Optimization Algorithm, which effectively identifies the optimal shape properties for feature selection. The final classification is executed using a Back Propagation Neural Network Classifier. The efficacy of our model was tested using images of barley afflicted with the specified diseases. The results were compelling, with the model achieving a remarkable accuracy rate, largely attributable to the integration of the grasshopper optimization algorithm in the feature selection stage.","PeriodicalId":13205,"journal":{"name":"IEEE Open Journal of the Computer Society","volume":"5 ","pages":"530-541"},"PeriodicalIF":0.0,"publicationDate":"2024-09-10","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10670315","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142208848","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Stitch-Able Split Learning Assisted Multi-UAV Systems","authors":"Tingkai Sun;Xiaoyan Wang;Xiucai Ye;Biao Han","doi":"10.1109/OJCS.2024.3447773","DOIUrl":"https://doi.org/10.1109/OJCS.2024.3447773","url":null,"abstract":"Unmanned aerial vehicles (UAVs), commonly known as drones, have gained widespread popularity due to their ease of deployment and high agility in various applications. In scenarios such as search missions and target tracking, conducting complex and computation-intensive tasks in multi-UAV systems have become essential. Recent investigations have explored the integration of collaborative centralized learning (CL) and federated learning (FL) into multi-UAV systems. However, CL methods raise privacy concerns and may suffer from communication delays, while FL methods demand high UAV-side computation capability. To address these challenges, split learning (SL) emerges as a promising alternative, offering reduced learning iteration time and improved accuracy in resource-constrained edge clients. In this study, we leverage SL and Stitch-able Neural Network (SN-NET) to propose a novel Stitch-able Split Learning (SSL) approach for multi-UAV systems. The proposed SSL approach is capable of tackling challenges in terms of device instability and model heterogeneity that associated in multi-UAV systems. Comparative simulations are conducted, evaluating its performance against CL, FL, traditional SL and SFLV1 (SplitFed Learning V1) approaches to establish its superiority.","PeriodicalId":13205,"journal":{"name":"IEEE Open Journal of the Computer Society","volume":"5 ","pages":"418-429"},"PeriodicalIF":0.0,"publicationDate":"2024-08-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10643654","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142152077","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}