Pub Date : 2025-12-01DOI: 10.1016/j.icte.2025.10.009
Tacettin Ayar, Deniz Turgay Altilar
Mobile end-users cannot utilize high upload bandwidths available in wireless access networks since wireless link based packet losses dramatically degrade TCP performance. We propose a TCP performance enhancing proxy pair called PROPER that detects wireless-based packet losses early and prevents TCP performance degradation. PROPER is transparent to TCP and requires no modifications on either TCP sender or TCP receiver. PROPER works in harmony with various TCP variants such as Reno, CUBIC, Veno and BBR. Netem-based performance and fairness emulation tests show that PROPER not only prevents TCP performance degradation on wireless access networks but also can safely coexist with regular TCP traffic.
{"title":"PROPER: A PROxy pair for uplink PERformance enhancement in wireless access networks","authors":"Tacettin Ayar, Deniz Turgay Altilar","doi":"10.1016/j.icte.2025.10.009","DOIUrl":"10.1016/j.icte.2025.10.009","url":null,"abstract":"<div><div>Mobile end-users cannot utilize high upload bandwidths available in wireless access networks since wireless link based packet losses dramatically degrade TCP performance. We propose a TCP performance enhancing proxy pair called PROPER that detects wireless-based packet losses early and prevents TCP performance degradation. PROPER is transparent to TCP and requires no modifications on either TCP sender or TCP receiver. PROPER works in harmony with various TCP variants such as Reno, CUBIC, Veno and BBR. Netem-based performance and fairness emulation tests show that PROPER not only prevents TCP performance degradation on wireless access networks but also can safely coexist with regular TCP traffic.</div></div>","PeriodicalId":48526,"journal":{"name":"ICT Express","volume":"11 6","pages":"Pages 1257-1264"},"PeriodicalIF":4.2,"publicationDate":"2025-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145705496","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}
Recent breakthroughs in generative reasoning have fundamentally reshaped how large language models (LLMs) address complex tasks, enabling them to dynamically retrieve, refine, and organize information into coherent, multi-step reasoning chains. Techniques such as inference-time scaling, reinforcement learning, supervised fine-tuning, and distillation have been effectively applied to state-of-the-art models, including DeepSeek-R1, OpenAI’s o1 and o3, GPT-4o, Qwen-32B, and various Llama variants, significantly enhancing their reasoning capabilities. In this paper, we present a comprehensive review of the top 27 LLMs released between 2023 and 2025, such as Mistral AI Small 3 24B, DeepSeek-R1, Search-o1, QwQ-32B, and Phi-4, and analyze their core innovations and performance improvements.
We also provide a detailed overview of recent advancements in multilingual large language models (MLLMs), emphasizing methods that improve cross-lingual reasoning and address the limitations of English-centric training. In parallel, we present a comprehensive review of progress in State Space Model (SSM)-based architectures, including models like Mamba, which demonstrate improved efficiency for long-context processing compared to Transformer-based approaches. Our analysis covers training strategies such as general optimization techniques, mixture-of-experts (MoE) configurations, retrieval-augmented generation (RAG), chain-of-thought prompting, self-improvement methods, and test-time compute scaling and distillation frameworks.
Finally, we identify key challenges for future research, including enabling multi-step reasoning without human supervision, improving robustness in chained task execution, balancing structured prompting with generative flexibility, and enhancing the integration of long-context retrieval and external tools.
生成推理的最新突破从根本上重塑了大型语言模型（llm）处理复杂任务的方式，使它们能够动态地检索、提炼和组织信息到连贯的、多步骤的推理链中。推理时间缩放、强化学习、监督微调和蒸馏等技术已有效应用于最先进的模型，包括DeepSeek-R1、OpenAI的o1和o3、gpt - 40、Qwen-32B和各种Llama变体，显著提高了它们的推理能力。在本文中，我们全面回顾了2023年至2025年间发布的27个顶级llm，如Mistral AI Small 3 24B、DeepSeek-R1、search - 01、QwQ-32B和Phi-4，并分析了它们的核心创新和性能改进。我们还详细概述了多语言大型语言模型（mllm）的最新进展，强调了改进跨语言推理和解决以英语为中心的培训局限性的方法。同时，我们对基于状态空间模型（SSM）的体系结构的进展进行了全面的回顾，包括像Mamba这样的模型，与基于transformer的方法相比，它证明了长上下文处理的效率提高。我们的分析涵盖了训练策略，例如一般优化技术、专家组合（MoE）配置、检索增强生成（RAG）、思维链提示、自我改进方法以及测试时间计算缩放和蒸馏框架。最后，我们确定了未来研究的关键挑战，包括在没有人类监督的情况下实现多步骤推理，提高链式任务执行的鲁棒性，平衡结构化提示与生成灵活性，以及增强长上下文检索和外部工具的集成。
{"title":"Reasoning beyond limits: Advances and open problems for LLMs","authors":"Mohamed Amine Ferrag , Norbert Tihanyi , Merouane Debbah","doi":"10.1016/j.icte.2025.09.003","DOIUrl":"10.1016/j.icte.2025.09.003","url":null,"abstract":"<div><div>Recent breakthroughs in generative reasoning have fundamentally reshaped how large language models (LLMs) address complex tasks, enabling them to dynamically retrieve, refine, and organize information into coherent, multi-step reasoning chains. Techniques such as inference-time scaling, reinforcement learning, supervised fine-tuning, and distillation have been effectively applied to state-of-the-art models, including DeepSeek-R1, OpenAI’s o1 and o3, GPT-4o, Qwen-32B, and various Llama variants, significantly enhancing their reasoning capabilities. In this paper, we present a comprehensive review of the top 27 LLMs released between 2023 and 2025, such as Mistral AI Small 3 24B, DeepSeek-R1, Search-o1, QwQ-32B, and Phi-4, and analyze their core innovations and performance improvements.</div><div>We also provide a detailed overview of recent advancements in multilingual large language models (MLLMs), emphasizing methods that improve cross-lingual reasoning and address the limitations of English-centric training. In parallel, we present a comprehensive review of progress in State Space Model (SSM)-based architectures, including models like Mamba, which demonstrate improved efficiency for long-context processing compared to Transformer-based approaches. Our analysis covers training strategies such as general optimization techniques, mixture-of-experts (MoE) configurations, retrieval-augmented generation (RAG), chain-of-thought prompting, self-improvement methods, and test-time compute scaling and distillation frameworks.</div><div>Finally, we identify key challenges for future research, including enabling multi-step reasoning without human supervision, improving robustness in chained task execution, balancing structured prompting with generative flexibility, and enhancing the integration of long-context retrieval and external tools.</div></div>","PeriodicalId":48526,"journal":{"name":"ICT Express","volume":"11 6","pages":"Pages 1054-1096"},"PeriodicalIF":4.2,"publicationDate":"2025-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145705499","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 : 2025-12-01DOI: 10.1016/j.icte.2025.11.009
Yamuna Tumma, Mahesh Miriyala
Reliable channel estimation is critical for achieving high-speed and energy-efficient communication in Unmanned Aerial Vehicle-Free Space Optical (UAV-FSO) systems, particularly under dynamic impairments such as atmospheric turbulence (AT) and pointing errors (PEs). This paper proposes a pilot-free channel estimation framework based on a Variational Autoencoder (VAE). The system employs Intensity Modulation/Direct Detection (IM/DD) with -ary one-hot encoded symbols (). The VAE encodes noisy received signals into a 128-dimensional latent space and reconstructs the transmitted data, effectively learning the joint effects of AT, PEs, and AWGN. Unlike prior works that primarily consider boresight or Gaussian-jitter PEs, this study explicitly incorporates a Nakagami-modeled PE distribution, capturing UAV-induced beam misalignment under mobility, vibration, and turbulence coupling. Simulation results show that the proposed VAE significantly outperforms conventional estimators (LS, MMSE, LMMSE) and deep learning baselines (AE, DNN, CNN) across various turbulence strengths. Under strong turbulence and PEs, the VAE attains nearly two-fold lower MSE compared to CNN and DNN. In addition, evaluation on real turbulence-impaired datasets further validates robustness and generalization. The proposed pilot-free scheme delivers accurate channel estimation, reduced BER, and improved spectral efficiency, making it suitable for real-time adaptive UAV-FSO communication.
{"title":"Deep learning-based pilot-free channel estimation of UAV-FSO system using variational auto-encoder","authors":"Yamuna Tumma, Mahesh Miriyala","doi":"10.1016/j.icte.2025.11.009","DOIUrl":"10.1016/j.icte.2025.11.009","url":null,"abstract":"<div><div>Reliable channel estimation is critical for achieving high-speed and energy-efficient communication in Unmanned Aerial Vehicle-Free Space Optical (UAV-FSO) systems, particularly under dynamic impairments such as atmospheric turbulence (AT) and pointing errors (PEs). This paper proposes a pilot-free channel estimation framework based on a Variational Autoencoder (VAE). The system employs Intensity Modulation/Direct Detection (IM/DD) with <span><math><mi>M</mi></math></span>-ary one-hot encoded symbols (<span><math><mrow><mi>M</mi><mo>=</mo><mn>16</mn></mrow></math></span>). The VAE encodes noisy received signals into a 128-dimensional latent space and reconstructs the transmitted data, effectively learning the joint effects of AT, PEs, and AWGN. Unlike prior works that primarily consider boresight or Gaussian-jitter PEs, this study explicitly incorporates a Nakagami-modeled PE distribution, capturing UAV-induced beam misalignment under mobility, vibration, and turbulence coupling. Simulation results show that the proposed VAE significantly outperforms conventional estimators (LS, MMSE, LMMSE) and deep learning baselines (AE, DNN, CNN) across various turbulence strengths. Under strong turbulence and PEs, the VAE attains nearly two-fold lower MSE compared to CNN and DNN. In addition, evaluation on real turbulence-impaired datasets further validates robustness and generalization. The proposed pilot-free scheme delivers accurate channel estimation, reduced BER, and improved spectral efficiency, making it suitable for real-time adaptive UAV-FSO communication.</div></div>","PeriodicalId":48526,"journal":{"name":"ICT Express","volume":"11 6","pages":"Pages 1162-1166"},"PeriodicalIF":4.2,"publicationDate":"2025-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145705608","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 : 2025-12-01DOI: 10.1016/j.icte.2025.10.003
Chen Zhang , Xiang Gui , Gourab Sen Gupta , Syed Faraz Hasan
This paper proposes a deterministic Global Optimization Algorithm (GOA) for UAV-assisted communications, developed as an enhancement to the benchmark Two-Stage Optimization Algorithm (TSOA). The algorithm simultaneously addresses the dual objectives of maximizing ground user (GU) coverage and minimizing total power consumption in multiple UAV systems. Unlike existing literature, which predominantly relies on heuristic approaches, GOA provides a more precise and systematic solution to achieve optimal performance. Comprehensive simulations demonstrate that GOA achieves a 3.68 % increase in coverage count versus SOA under clustered GU distributions while delivering energy savings approximately 2.47 % (uniform) and 2.6 % (clustered) relative to the TSOA benchmark. Crucially, these efficiency gains are realized while maintaining superior GU coverage maximization versus all benchmarked methods. Both numerical results and visual analyses conclusively validate the proposed algorithm's outperformance of existing benchmarks.
Pub Date : 2025-12-01DOI: 10.1016/j.icte.2025.11.003
Seonjoo Choi , Hoki Baek , Jaesung Lim
Beamforming training (BFT) is a critical process for establishing directional links in millimeter-wave (mmWave) wireless local area networks. However, its performance significantly degrades due to frequent slot collisions. This paper proposes a non-orthogonal multiple access (NOMA)-based exhaustive search (NES) scheme, which allows multiple stations (STAs) to perform BFT concurrently by independently selecting predefined transmit power levels. We integrated the NOMA-based technique into a time-based beam collision avoidance (TBCA) scheme, named NOMA-TBCA (NTBCA). The proposed NES and NTBCA schemes are analytically modeled and evaluated through simulations. The results confirm that the combined approach effectively enhances the association probability and the throughput.
{"title":"Efficient beamforming training scheme using NOMA in mmWave WLANs","authors":"Seonjoo Choi , Hoki Baek , Jaesung Lim","doi":"10.1016/j.icte.2025.11.003","DOIUrl":"10.1016/j.icte.2025.11.003","url":null,"abstract":"<div><div>Beamforming training (BFT) is a critical process for establishing directional links in millimeter-wave (mmWave) wireless local area networks. However, its performance significantly degrades due to frequent slot collisions. This paper proposes a non-orthogonal multiple access (NOMA)-based exhaustive search (NES) scheme, which allows multiple stations (STAs) to perform BFT concurrently by independently selecting predefined transmit power levels. We integrated the NOMA-based technique into a time-based beam collision avoidance (TBCA) scheme, named NOMA-TBCA (NTBCA). The proposed NES and NTBCA schemes are analytically modeled and evaluated through simulations. The results confirm that the combined approach effectively enhances the association probability and the throughput.</div></div>","PeriodicalId":48526,"journal":{"name":"ICT Express","volume":"11 6","pages":"Pages 1286-1290"},"PeriodicalIF":4.2,"publicationDate":"2025-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145705501","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 : 2025-12-01DOI: 10.1016/j.icte.2025.09.012
Mahalinoro Razafimanjato, Malik Muhammad Saad, Dongkyun Kim
The Internet of Vehicles (IoV), a critical component of Intelligent Transportation Systems (ITS), enhances driving safety and traffic efficiency through real-time data exchange. However, the dynamic and heterogeneous nature of IoV introduces significant security and trust challenges. To address these, trust management systems have emerged as vital mechanisms to ensure the reliability and integrity of data exchanged between vehicles. Blockchain technology offers a robust framework for addressing security and trust issues in IoV environments. The decentralized, tamper-resistant, and transparent nature of the blockchain makes it suitable for complex vehicular environments. This survey provides an overview of state-of-the-art blockchain-based trust management systems in IoV. Following a systematic literature review that filtered 8,280 publications to 63 core studies from 2019 to 2024, we present a thematic classification of existing solutions, focusing on those employing public and private blockchains. Unlike previous surveys, our work focuses specifically on the intersection of blockchain and trust management systems in IoV by analyzing approaches across four dimensions: trust computation methods, such as game theory and AI-driven models; blockchain scaling solutions, including sharding, sidechains, and optimized consensus mechanisms; integration with emerging technologies such as 5G/6G, Digital Twins, and Federated Learning; and security and privacy mechanisms. Finally, this survey identifies current challenges and provides future research directions, highlighting the need for more scalable, adaptive, secure, and privacy-preserving trust management systems in IoV.
{"title":"Blockchain-based trust management systems in the Internet of Vehicles: A comprehensive survey","authors":"Mahalinoro Razafimanjato, Malik Muhammad Saad, Dongkyun Kim","doi":"10.1016/j.icte.2025.09.012","DOIUrl":"10.1016/j.icte.2025.09.012","url":null,"abstract":"<div><div>The Internet of Vehicles (IoV), a critical component of Intelligent Transportation Systems (ITS), enhances driving safety and traffic efficiency through real-time data exchange. However, the dynamic and heterogeneous nature of IoV introduces significant security and trust challenges. To address these, trust management systems have emerged as vital mechanisms to ensure the reliability and integrity of data exchanged between vehicles. Blockchain technology offers a robust framework for addressing security and trust issues in IoV environments. The decentralized, tamper-resistant, and transparent nature of the blockchain makes it suitable for complex vehicular environments. This survey provides an overview of state-of-the-art blockchain-based trust management systems in IoV. Following a systematic literature review that filtered 8,280 publications to 63 core studies from 2019 to 2024, we present a thematic classification of existing solutions, focusing on those employing public and private blockchains. Unlike previous surveys, our work focuses specifically on the intersection of blockchain and trust management systems in IoV by analyzing approaches across four dimensions: trust computation methods, such as game theory and AI-driven models; blockchain scaling solutions, including sharding, sidechains, and optimized consensus mechanisms; integration with emerging technologies such as 5G/6G, Digital Twins, and Federated Learning; and security and privacy mechanisms. Finally, this survey identifies current challenges and provides future research directions, highlighting the need for more scalable, adaptive, secure, and privacy-preserving trust management systems in IoV.</div></div>","PeriodicalId":48526,"journal":{"name":"ICT Express","volume":"11 6","pages":"Pages 1265-1285"},"PeriodicalIF":4.2,"publicationDate":"2025-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145705497","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 : 2025-12-01DOI: 10.1016/j.icte.2025.09.001
Kshiteesh Mani, Ajitha K.B. Shenoy
The rapid growth of web applications has increased the need for advanced features and strong security. Artificial intelligence (AI) and machine learning (ML) models play a crucial role in meeting these needs by improving efficiency and enhancing security. However, integrating these models into web applications can be challenging due to complex implementation and potential security risks. This paper compares Python and Node.js, two popular technology stacks, to determine their effectiveness in integrating ML models into web applications. It also explores the role of web application firewalls (WAF) and the ML algorithms that support them, analyzing current trends in their use and adoption. The overarching objective is to discern the technology stack that provides superior support for back-end ML integration and to identify the ML algorithms that are most effective in enhancing WAF capabilities against sophisticated security threats. By offering a synthesis of technical and security insights, this research seeks to empower developers and cybersecurity practitioners with the knowledge required to make well-informed decisions regarding technology stack selection and the implementation of ML-driven security mechanisms in web application development.
{"title":"Machine learning models in web applications: A comprehensive review","authors":"Kshiteesh Mani, Ajitha K.B. Shenoy","doi":"10.1016/j.icte.2025.09.001","DOIUrl":"10.1016/j.icte.2025.09.001","url":null,"abstract":"<div><div>The rapid growth of web applications has increased the need for advanced features and strong security. Artificial intelligence (AI) and machine learning (ML) models play a crucial role in meeting these needs by improving efficiency and enhancing security. However, integrating these models into web applications can be challenging due to complex implementation and potential security risks. This paper compares Python and Node.js, two popular technology stacks, to determine their effectiveness in integrating ML models into web applications. It also explores the role of web application firewalls (WAF) and the ML algorithms that support them, analyzing current trends in their use and adoption. The overarching objective is to discern the technology stack that provides superior support for back-end ML integration and to identify the ML algorithms that are most effective in enhancing WAF capabilities against sophisticated security threats. By offering a synthesis of technical and security insights, this research seeks to empower developers and cybersecurity practitioners with the knowledge required to make well-informed decisions regarding technology stack selection and the implementation of ML-driven security mechanisms in web application development.</div></div>","PeriodicalId":48526,"journal":{"name":"ICT Express","volume":"11 6","pages":"Pages 1110-1119"},"PeriodicalIF":4.2,"publicationDate":"2025-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145705603","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 : 2025-12-01DOI: 10.1016/j.icte.2025.09.008
Kristian Adi Nugraha , Igi Ardiyanto , Sunu Wibirama
Saliency prediction models are typically trained on natural images, focusing on features such as shape and color. However, predicting saliency in images with text is challenging because the human brain processes text differently than it processes visual objects. To address this research gap, we fine-tuned a saliency model to improve the accuracy of images containing text, specifically, movie posters. Our fine-tuned model — based on GSGNet and TranSalNet — outperformed the original models in predicting the saliency map for movie posters. The experimental results indicate that text elements exhibit patterns that can be learned for better saliency prediction.
{"title":"Fine-tuning deep neural network for saliency prediction in movie poster documents","authors":"Kristian Adi Nugraha , Igi Ardiyanto , Sunu Wibirama","doi":"10.1016/j.icte.2025.09.008","DOIUrl":"10.1016/j.icte.2025.09.008","url":null,"abstract":"<div><div>Saliency prediction models are typically trained on natural images, focusing on features such as shape and color. However, predicting saliency in images with text is challenging because the human brain processes text differently than it processes visual objects. To address this research gap, we fine-tuned a saliency model to improve the accuracy of images containing text, specifically, movie posters. Our fine-tuned model — based on GSGNet and TranSalNet — outperformed the original models in predicting the saliency map for movie posters. The experimental results indicate that text elements exhibit patterns that can be learned for better saliency prediction.</div></div>","PeriodicalId":48526,"journal":{"name":"ICT Express","volume":"11 6","pages":"Pages 1181-1185"},"PeriodicalIF":4.2,"publicationDate":"2025-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145705490","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 : 2025-12-01DOI: 10.1016/j.icte.2025.11.007
Seonghun Hong , Donghyun Lee , Dongwook Won , Wonjong Noh , Sungrae Cho
Semantic communication shifts the focus from bit-level accuracy to task-relevant meaning. However, most methods assume equal importance across semantic units and rely on costly retraining, limiting scalability. This work proposes an importance-aware framework that accounts for unequal feature contributions. It introduces a new metric, importance-weighted semantic spectral efficiency (wSSE), to prioritize task-relevant features. It develops an empirically derived feature-accuracy matrix, inspired by saturation behavior. The framework enables importance-aware feature selection and subchannel allocation without online learning. This framework targets resource-constrained edge systems such as IoT cameras, UAV detection, and AR/VR. Experiments show up to 73.3% higher efficiency under constrained resources.
{"title":"Feature-importance-aware transmission control in semantic communications","authors":"Seonghun Hong , Donghyun Lee , Dongwook Won , Wonjong Noh , Sungrae Cho","doi":"10.1016/j.icte.2025.11.007","DOIUrl":"10.1016/j.icte.2025.11.007","url":null,"abstract":"<div><div>Semantic communication shifts the focus from bit-level accuracy to task-relevant meaning. However, most methods assume equal importance across semantic units and rely on costly retraining, limiting scalability. This work proposes an importance-aware framework that accounts for unequal feature contributions. It introduces a new metric, importance-weighted semantic spectral efficiency (wSSE), to prioritize task-relevant features. It develops an empirically derived feature-accuracy matrix, inspired by saturation behavior. The framework enables importance-aware feature selection and subchannel allocation without online learning. This framework targets resource-constrained edge systems such as IoT cameras, UAV detection, and AR/VR. Experiments show up to 73.3% higher efficiency under constrained resources.</div></div>","PeriodicalId":48526,"journal":{"name":"ICT Express","volume":"11 6","pages":"Pages 1015-1020"},"PeriodicalIF":4.2,"publicationDate":"2025-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145705599","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 : 2025-12-01DOI: 10.1016/j.icte.2025.11.006
Rafat Bin Mofidul, Al Imran, Md. Shahriar Nazim, Yeong Min Jang
Optical Intelligent Reflective Surface (OIRS)-assisted MIMO VLC offers high-capacity communication but struggles with accurate channel estimation due to multipath and noise. We propose a hybrid CNN–Swin Transformer denoising network (HCSTNet), integrating a CNN-based noise estimator with a Swin Transformer bottleneck to learn local features and global spatial dependencies jointly. In a realistic indoor MIMO VLC setup with LOS, NLOS, and OIRS links, HCSTNet achieves an NMSE of – and outperforms existing models in terms of PSNR while also reducing the parameter count and inference time. These results demonstrate the efficiency and robustness of HCSTNet for practical VLC applications.
{"title":"Hybrid CNN-Swin Transformer denoising network for channel estimation in Optical IRS-assisted indoor MIMO VLC system","authors":"Rafat Bin Mofidul, Al Imran, Md. Shahriar Nazim, Yeong Min Jang","doi":"10.1016/j.icte.2025.11.006","DOIUrl":"10.1016/j.icte.2025.11.006","url":null,"abstract":"<div><div>Optical Intelligent Reflective Surface (OIRS)-assisted MIMO VLC offers high-capacity communication but struggles with accurate channel estimation due to multipath and noise. We propose a hybrid CNN–Swin Transformer denoising network (HCSTNet), integrating a CNN-based noise estimator with a Swin Transformer bottleneck to learn local features and global spatial dependencies jointly. In a realistic indoor MIMO VLC setup with LOS, NLOS, and OIRS links, HCSTNet achieves an NMSE of <span><math><mrow><mn>1</mn><msup><mrow><mn>0</mn></mrow><mrow><mo>−</mo><mn>3</mn></mrow></msup></mrow></math></span>–<span><math><mrow><mn>1</mn><msup><mrow><mn>0</mn></mrow><mrow><mo>−</mo><mn>4</mn></mrow></msup></mrow></math></span> and outperforms existing models in terms of PSNR while also reducing the parameter count and inference time. These results demonstrate the efficiency and robustness of HCSTNet for practical VLC applications.</div></div>","PeriodicalId":48526,"journal":{"name":"ICT Express","volume":"11 6","pages":"Pages 1047-1053"},"PeriodicalIF":4.2,"publicationDate":"2025-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145705602","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}
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