Pub Date : 2026-02-01Epub Date: 2025-12-11DOI: 10.1016/j.icte.2025.12.003
Abdul Wahid Soomro , Miss Laiha Mat Kiah , Rafidah Md Noor , Salim Newaz Kazi , Kaleemullah Shaikh , Wajahat Ahmed Khan , Ihsan Ali
Fouling in heat exchangers (HXs) affects various industries by lowering efficiency and increasing costs. Traditional fouling-prediction models often do not reflect important mechanistic information and thus become very complex and less reliable. The applications of artificial intelligence (AI), machine learning (ML), and deep learning (DL) open new frontiers, as these techniques can model complex correlations and work with large volumes of data. This review synthesizes 51 articles published between 2005 and June 2025, outlining key trends, persistent research limitations, and emerging directions. Models such as artificial neural networks (ANNs)/deep neural networks (DNNs) and Gaussian process regression (GPR) deliver the optimal results in terms of accurate prediction.
{"title":"Artificial intelligence in industrial heat exchanger fouling prediction: A 20-year systematic review of AI, ML, and DL approaches","authors":"Abdul Wahid Soomro , Miss Laiha Mat Kiah , Rafidah Md Noor , Salim Newaz Kazi , Kaleemullah Shaikh , Wajahat Ahmed Khan , Ihsan Ali","doi":"10.1016/j.icte.2025.12.003","DOIUrl":"10.1016/j.icte.2025.12.003","url":null,"abstract":"<div><div>Fouling in heat exchangers (HXs) affects various industries by lowering efficiency and increasing costs. Traditional fouling-prediction models often do not reflect important mechanistic information and thus become very complex and less reliable. The applications of artificial intelligence (AI), machine learning (ML), and deep learning (DL) open new frontiers, as these techniques can model complex correlations and work with large volumes of data. This review synthesizes 51 articles published between 2005 and June 2025, outlining key trends, persistent research limitations, and emerging directions. Models such as artificial neural networks (ANNs)/deep neural networks (DNNs) and Gaussian process regression (GPR) deliver the optimal results in terms of accurate prediction.</div></div>","PeriodicalId":48526,"journal":{"name":"ICT Express","volume":"12 1","pages":"Pages 92-110"},"PeriodicalIF":4.2,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146154373","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-01Epub Date: 2025-12-17DOI: 10.1016/j.icte.2025.12.006
Sun-Jin Lee, So-Eun Jeon, Il-Gu Lee
Advances in next-generation wireless network technology is boosting the demand for ultra-low latency and reliable data-transmission technologies. Particularly, industries are adopting robust end-to-end encryption to enhance security. However, the delays inherent in conventional encryption methods make them unsuitable for meeting these demands. Therefore, this paper presents the Shamir secret sharing–partial encryption (SSS–PE) method, which encrypts only the minimal secret fragments within an SSS environment wherein data are distributed and transmitted. Experimental results indicate that SSS–PE improves secrecy throughput by 1.86 times and reduces latency by 1.38 times compared to the Advanced Encryption Standard.
{"title":"Partial encryption-based Shamir secret sharing for low-latency and secure networks","authors":"Sun-Jin Lee, So-Eun Jeon, Il-Gu Lee","doi":"10.1016/j.icte.2025.12.006","DOIUrl":"10.1016/j.icte.2025.12.006","url":null,"abstract":"<div><div>Advances in next-generation wireless network technology is boosting the demand for ultra-low latency and reliable data-transmission technologies. Particularly, industries are adopting robust end-to-end encryption to enhance security. However, the delays inherent in conventional encryption methods make them unsuitable for meeting these demands. Therefore, this paper presents the Shamir secret sharing–partial encryption (SSS–PE) method, which encrypts only the minimal secret fragments within an SSS environment wherein data are distributed and transmitted. Experimental results indicate that SSS–PE improves secrecy throughput by 1.86 times and reduces latency by 1.38 times compared to the Advanced Encryption Standard.</div></div>","PeriodicalId":48526,"journal":{"name":"ICT Express","volume":"12 1","pages":"Pages 168-174"},"PeriodicalIF":4.2,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146154378","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-01Epub Date: 2025-11-28DOI: 10.1016/j.icte.2025.11.016
Hyunwoo Jung , Jung-Bin Kim
This paper introduces a novel selective channel inversion (SCI) protocol for over-the-air computation (AirComp) networks. The proposed SCI protocol reduces the overhead associated with channel state information (CSI) by broadcasting a predefined channel threshold. Furthermore, it improves mean squared error (MSE) performance by permitting transmission only from sensor nodes with favorable channel conditions. For an arbitrary channel distribution, an exact closed-form expression for the MSE is derived, from which an asymptotic expression for a large number of sensor nodes is obtained. Numerical results demonstrate that the asymptotic expression closely matches the exact result, even for a relatively small number of sensor nodes. An optimization problem is formulated to determine the optimal channel threshold under Nakagami- fading channels, and it is analytically proven that the problem is convex. An algorithm for dynamically determining the threshold under arbitrary channel distributions is also presented. Numerical results demonstrate that SCI AirComp outperforms CI AirComp in scenarios with low signal-to-noise ratio (SNR), high fading severity, and a large number of sensor nodes.
2018 The Korean Institute of Communications and Information Sciences. Publishing Services by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
{"title":"Selective channel inversion protocol for over-the-air computation","authors":"Hyunwoo Jung , Jung-Bin Kim","doi":"10.1016/j.icte.2025.11.016","DOIUrl":"10.1016/j.icte.2025.11.016","url":null,"abstract":"<div><div>This paper introduces a novel selective channel inversion (SCI) protocol for over-the-air computation (AirComp) networks. The proposed SCI protocol reduces the overhead associated with channel state information (CSI) by broadcasting a predefined channel threshold. Furthermore, it improves mean squared error (MSE) performance by permitting transmission only from sensor nodes with favorable channel conditions. For an arbitrary channel distribution, an exact closed-form expression for the MSE is derived, from which an asymptotic expression for a large number of sensor nodes is obtained. Numerical results demonstrate that the asymptotic expression closely matches the exact result, even for a relatively small number of sensor nodes. An optimization problem is formulated to determine the optimal channel threshold under Nakagami-<span><math><mi>m</mi></math></span> fading channels, and it is analytically proven that the problem is convex. An algorithm for dynamically determining the threshold under arbitrary channel distributions is also presented. Numerical results demonstrate that SCI AirComp outperforms CI AirComp in scenarios with low signal-to-noise ratio (SNR), high fading severity, and a large number of sensor nodes.</div><div>2018 The Korean Institute of Communications and Information Sciences. Publishing Services by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (<span><span>http://creativecommons.org/licenses/by-nc-nd/4.0/</span><svg><path></path></svg></span>).</div></div>","PeriodicalId":48526,"journal":{"name":"ICT Express","volume":"12 1","pages":"Pages 6-12"},"PeriodicalIF":4.2,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146154420","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-01Epub Date: 2025-09-05DOI: 10.1016/j.icte.2025.09.002
Hongqing Wang , Jun Kit Chaw , Marizuana Mat Daud , Liantao Shi , Nannan Huang , Tin Tin Ting , Liuzhen Pu
Vehicle detection in complex traffic scenes remains challenging due to frequent occlusions, lighting variations, and extreme weather. We present Hybrid-YOLO, a real-time detection framework that unifies Mamba-based state space modeling, Transformer-driven global attention, and multi-scale feature fusion to achieve high accuracy at low computational cost. At its core, Hybrid-YOLO introduces a Dynamic Residual Stem (DR Stem) for adaptive feature calibration, a Hexa-Scan Selective Block (HSSBlock) for six-directional structural perception, and a Selective State Space Model (SSM) for efficient long-range dependency modeling. A Cross-Stage Scales Feature Extraction (CSSFE) module enriches spatial semantics for small-object detection, while a Sparse-Queries Cascade Self-Attention (SCS) module focuses computation on informative regions, enhancing robustness to clutter and background noise. Extensive experiments on KITTI, BDD100K, and IITM-HeTra show that Hybrid-YOLO achieves 90.11 [email protected] at 66.3 FPS, surpassing state-of-the-art methods in both accuracy and efficiency, and offering a promising solution for real-world intelligent transportation systems.
{"title":"Hybrid-YOLO: Lightweight Mamba-Transformer Hybrid with multi-scale fusion for real-world traffic detection","authors":"Hongqing Wang , Jun Kit Chaw , Marizuana Mat Daud , Liantao Shi , Nannan Huang , Tin Tin Ting , Liuzhen Pu","doi":"10.1016/j.icte.2025.09.002","DOIUrl":"10.1016/j.icte.2025.09.002","url":null,"abstract":"<div><div>Vehicle detection in complex traffic scenes remains challenging due to frequent occlusions, lighting variations, and extreme weather. We present Hybrid-YOLO, a real-time detection framework that unifies Mamba-based state space modeling, Transformer-driven global attention, and multi-scale feature fusion to achieve high accuracy at low computational cost. At its core, Hybrid-YOLO introduces a Dynamic Residual Stem (DR Stem) for adaptive feature calibration, a Hexa-Scan Selective Block (HSSBlock) for six-directional structural perception, and a Selective State Space Model (SSM) for efficient long-range dependency modeling. A Cross-Stage Scales Feature Extraction (CSSFE) module enriches spatial semantics for small-object detection, while a Sparse-Queries Cascade Self-Attention (SCS) module focuses computation on informative regions, enhancing robustness to clutter and background noise. Extensive experiments on KITTI, BDD100K, and IITM-HeTra show that Hybrid-YOLO achieves 90.11 [email protected] at 66.3 FPS, surpassing state-of-the-art methods in both accuracy and efficiency, and offering a promising solution for real-world intelligent transportation systems.</div></div>","PeriodicalId":48526,"journal":{"name":"ICT Express","volume":"12 1","pages":"Pages 214-222"},"PeriodicalIF":4.2,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146154412","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-01Epub Date: 2025-12-18DOI: 10.1016/j.icte.2025.12.005
Kichang Lee , JeongGil Ko , Songkuk Kim , JaeYeon Park
On-device machine learning is fundamentally constrained by limited storage, especially in continuous data collection scenarios where sensor or vision streams accumulate rapidly. This paper empirically investigates storage-aware learning, characterizing the trade-off between data quantity and data quality under lossy compression. Using the CIFAR-10 dataset as a controlled benchmark, we systematically vary both the amount and the fidelity of training data to understand their joint impact on model performance. Our results reveal that (1) neither maximizing quantity nor quality alone yields optimal accuracy, emphasizing that the optimal trade-off between them depends nonlinearly on the available storage budget, and (2) data samples exhibit differential sensitivity to compression, motivating a sample-wise adaptive compression policy. These findings challenge uniform data-retention strategies such as naive data dropping or fixed-rate compression, and establish a foundation for adaptive, storage-efficient learning systems on resource-limited devices. This work opens new directions toward generalizable, storage-aware on-device intelligence.
{"title":"Toward storage-aware learning with compressed data an empirical exploratory study on JPEG","authors":"Kichang Lee , JeongGil Ko , Songkuk Kim , JaeYeon Park","doi":"10.1016/j.icte.2025.12.005","DOIUrl":"10.1016/j.icte.2025.12.005","url":null,"abstract":"<div><div>On-device machine learning is fundamentally constrained by limited storage, especially in continuous data collection scenarios where sensor or vision streams accumulate rapidly. This paper empirically investigates storage-aware learning, characterizing the trade-off between data quantity and data quality under lossy compression. Using the CIFAR-10 dataset as a controlled benchmark, we systematically vary both the amount and the fidelity of training data to understand their joint impact on model performance. Our results reveal that (1) neither maximizing quantity nor quality alone yields optimal accuracy, emphasizing that the optimal trade-off between them depends nonlinearly on the available storage budget, and (2) data samples exhibit differential sensitivity to compression, motivating a sample-wise adaptive compression policy. These findings challenge uniform data-retention strategies such as naive data dropping or fixed-rate compression, and establish a foundation for adaptive, storage-efficient learning systems on resource-limited devices. This work opens new directions toward generalizable, storage-aware on-device intelligence.</div></div>","PeriodicalId":48526,"journal":{"name":"ICT Express","volume":"12 1","pages":"Pages 50-54"},"PeriodicalIF":4.2,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146154369","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Stock price forecasting is a crucial challenge in FinTech industries, with implications that extend to algorithmic trading. In recent studies, Graph Neural Networks (GNNs) are employed for prediction; however, they are still limited to involving macroeconomic contexts. Hence, in this study, a novel GNN-based method for stock price forecasting is designed, with a graph building structure influenced by macroeconomic variables, namely inflation, interest rate, and GDP growth regimes. Our model captures the relationships between stocks on the basis of regime-specific, macro-driven static graphs along with an LSTM model. The proposed approach outperforms existing methods and provides a new viewpoint on stock forecasting.
{"title":"Macroeconomic context-aware graph topology learning for stock price forecasting using graph neural network","authors":"Amna Sarwar , Fizza Bukhari , Asma Sattar , Maryam Bukhari , Zahoor ur Rehman , Sungwoo Park , Seungmin Rho","doi":"10.1016/j.icte.2025.12.002","DOIUrl":"10.1016/j.icte.2025.12.002","url":null,"abstract":"<div><div>Stock price forecasting is a crucial challenge in FinTech industries, with implications that extend to algorithmic trading. In recent studies, Graph Neural Networks (GNNs) are employed for prediction; however, they are still limited to involving macroeconomic contexts. Hence, in this study, a novel GNN-based method for stock price forecasting is designed, with a graph building structure influenced by macroeconomic variables, namely inflation, interest rate, and GDP growth regimes. Our model captures the relationships between stocks on the basis of regime-specific, macro-driven static graphs along with an LSTM model. The proposed approach outperforms existing methods and provides a new viewpoint on stock forecasting.</div></div>","PeriodicalId":48526,"journal":{"name":"ICT Express","volume":"12 1","pages":"Pages 13-19"},"PeriodicalIF":4.2,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146154421","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-01Epub Date: 2025-12-29DOI: 10.1016/j.icte.2025.12.013
Thien Huynh-The , Toan Van Nguyen , Phuong Luu Vo , Huu-Tai Nguyen
We introduce LDSeCom, a novel lightweight-to-diffusion framework for semantic image communication. LDSeCom addresses bandwidth constraints by developing LSNet, a lightweight, loop-based segmentation model at the sender, and an improved diffusion model guided by our AFM-Net at the receiver. LSNet efficiently compresses images into semantic maps, while AFM-Net’s adaptive feature modulation ensures high-quality image reconstruction. On benchmark datasets, our LSNet achieves competitive accuracy with only 0.5M parameters, while our diffusion model improves image reconstruction quality by up to 28.51% mFID. The framework enables high-fidelity results from semantic maps compressed to of the original size, proving its efficiency for bandwidth-constrained scenarios.
{"title":"A lightweight-to-diffusion framework for semantic image communications","authors":"Thien Huynh-The , Toan Van Nguyen , Phuong Luu Vo , Huu-Tai Nguyen","doi":"10.1016/j.icte.2025.12.013","DOIUrl":"10.1016/j.icte.2025.12.013","url":null,"abstract":"<div><div>We introduce LDSeCom, a novel lightweight-to-diffusion framework for semantic image communication. LDSeCom addresses bandwidth constraints by developing LSNet, a lightweight, loop-based segmentation model at the sender, and an improved diffusion model guided by our AFM-Net at the receiver. LSNet efficiently compresses images into semantic maps, while AFM-Net’s adaptive feature modulation ensures high-quality image reconstruction. On benchmark datasets, our LSNet achieves competitive accuracy with only 0.5M parameters, while our diffusion model improves image reconstruction quality by up to 28.51% mFID. The framework enables high-fidelity results from semantic maps compressed to <span><math><mrow><mn>1</mn><mo>/</mo><mn>80</mn></mrow></math></span> of the original size, proving its efficiency for bandwidth-constrained scenarios.</div></div>","PeriodicalId":48526,"journal":{"name":"ICT Express","volume":"12 1","pages":"Pages 175-179"},"PeriodicalIF":4.2,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146154379","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-01Epub Date: 2025-11-29DOI: 10.1016/j.icte.2025.11.015
Sehwan Heo , Junbeom Moon , Soon Ki Jung
Action recognition has emerged as a central research problem in computer vision, aiming to identify and understand human actions from video data. Over the past decade, the field has advanced from early convolutional approaches to sophisticated architectures capable of capturing complex spatio-temporal dependencies. This survey provides a comprehensive overview of action recognition across six major tasks: Action Classification, Temporal Action Localization, Spatio-temporal Action Localization, Temporal Action Segmentation, Online Action Detection, and Action Anticipation. For each task, we trace the methodological evolution from foundational models to recent state-of-the-art approaches, highlighting how key challenges such as long-range temporal modeling, viewpoint variation, boundary precision, over-segmentation, real-time inference, and future uncertainty have been addressed. We also reorganize benchmark results and evaluation metrics, presenting a unified perspective that facilitates fair comparisons and reproducible research. In addition, we analyze representative datasets, ranging from early benchmarks like UCF101 and HMDB51 to large-scale collections such as Kinetics, ActivityNet, and Epic-Kitchens, which have enabled rapid progress in both supervised and self-supervised learning. We discuss open issues and unresolved challenges, including the use of State Space Models for efficient sequence modeling, multimodal fusion techniques that dynamically assess modality reliability, synthetic data and weak supervision for reducing annotation costs, and fairness-aware frameworks that ensure ethical applicability. By consolidating a decade of progress, this survey offers a structured understanding of the action recognition landscape and aims to inspire further research toward robust, scalable, and responsible video understanding systems.
{"title":"Action recognition: A comprehensive survey of tasks, methods, and challenges","authors":"Sehwan Heo , Junbeom Moon , Soon Ki Jung","doi":"10.1016/j.icte.2025.11.015","DOIUrl":"10.1016/j.icte.2025.11.015","url":null,"abstract":"<div><div>Action recognition has emerged as a central research problem in computer vision, aiming to identify and understand human actions from video data. Over the past decade, the field has advanced from early convolutional approaches to sophisticated architectures capable of capturing complex spatio-temporal dependencies. This survey provides a comprehensive overview of action recognition across six major tasks: Action Classification, Temporal Action Localization, Spatio-temporal Action Localization, Temporal Action Segmentation, Online Action Detection, and Action Anticipation. For each task, we trace the methodological evolution from foundational models to recent state-of-the-art approaches, highlighting how key challenges such as long-range temporal modeling, viewpoint variation, boundary precision, over-segmentation, real-time inference, and future uncertainty have been addressed. We also reorganize benchmark results and evaluation metrics, presenting a unified perspective that facilitates fair comparisons and reproducible research. In addition, we analyze representative datasets, ranging from early benchmarks like UCF101 and HMDB51 to large-scale collections such as Kinetics, ActivityNet, and Epic-Kitchens, which have enabled rapid progress in both supervised and self-supervised learning. We discuss open issues and unresolved challenges, including the use of State Space Models for efficient sequence modeling, multimodal fusion techniques that dynamically assess modality reliability, synthetic data and weak supervision for reducing annotation costs, and fairness-aware frameworks that ensure ethical applicability. By consolidating a decade of progress, this survey offers a structured understanding of the action recognition landscape and aims to inspire further research toward robust, scalable, and responsible video understanding systems.</div></div>","PeriodicalId":48526,"journal":{"name":"ICT Express","volume":"12 1","pages":"Pages 32-49"},"PeriodicalIF":4.2,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146154367","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-01Epub Date: 2025-07-10DOI: 10.1016/j.icte.2025.06.016
Sangwoo Choi , Minsik Kim , Daeyoung Park
Federated learning (FL) enables distributed model training while preserving privacy, but frequent updates from many devices create substantial communication challenges. Over-the-air computation (AirComp) offers a solution by aggregating updates directly over wireless channels through signal superposition, reducing overhead. However, AirComp can increase the mean squared error (MSE) of aggregated signals, affecting model accuracy. This paper introduces a beamforming optimization framework for AirComp in FL systems, jointly optimizing base station beamforming and device transmission scaling to minimize MSE. Two algorithms are developed: a high-performance convex method (Miso-CVX) and a lower-complexity subgradient method (Miso-Subgradient), both balancing signal misalignment and noise. Extensive simulations show improved aggregation accuracy, convergence speed, and robustness to channel variations.
{"title":"Optimal beamforming in over-the-air federated learning for efficient model aggregation","authors":"Sangwoo Choi , Minsik Kim , Daeyoung Park","doi":"10.1016/j.icte.2025.06.016","DOIUrl":"10.1016/j.icte.2025.06.016","url":null,"abstract":"<div><div>Federated learning (FL) enables distributed model training while preserving privacy, but frequent updates from many devices create substantial communication challenges. Over-the-air computation (AirComp) offers a solution by aggregating updates directly over wireless channels through signal superposition, reducing overhead. However, AirComp can increase the mean squared error (MSE) of aggregated signals, affecting model accuracy. This paper introduces a beamforming optimization framework for AirComp in FL systems, jointly optimizing base station beamforming and device transmission scaling to minimize MSE. Two algorithms are developed: a high-performance convex method (Miso-CVX) and a lower-complexity subgradient method (Miso-Subgradient), both balancing signal misalignment and noise. Extensive simulations show improved aggregation accuracy, convergence speed, and robustness to channel variations.</div></div>","PeriodicalId":48526,"journal":{"name":"ICT Express","volume":"12 1","pages":"Pages 136-141"},"PeriodicalIF":4.2,"publicationDate":"2026-02-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146154375","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2026-02-01Epub Date: 2025-11-28DOI: 10.1016/j.icte.2025.11.017
Seungseok Sin , Sangmi Moon , Cheol Hong Kim , Intae Hwang
Accurate localization is essential for next-generation wireless systems. Traditional millimeter-wave (mmWave) techniques rely heavily on line-of-sight (LOS) paths, which limits their performance in non-line-of-sight (NLOS) environments. To overcome this challenge, we propose a non-terrestrial network (NTN) framework that employs an unmanned aerial vehicle–mounted reconfigurable intelligent surface (UAV-RIS) in conjunction with a Transformer-based refinement model. Unlike conventional regression or filtering approaches, the Transformer leverages self-attention mechanisms to refine coarse geometric estimates. Simulations using the DeepMIMO dataset show that more than 90% of users achieve sub-meter localization accuracy, representing a 35% improvement over existing baselines. These results demonstrate the novelty and effectiveness of integrating RIS adaptability with Transformer-based learning to enable robust, high-precision localization.
扫码关注我们
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号