IF 3.6 3区计算机科学

IEEE Access Pub Date : 2026-03-17 eCollection Date: 2026-01-01 DOI: 10.1109/ACCESS.2026.3674701

Kshitij Kumar, Kaustubh Deshpande, Naveen Kalur, Garima Chauhan, Deepti Chugh, Subramaniam Ganesh, Arjun Ramakrishnan

{"title":"A Translational Platform for Polyimide Neural Interfaces: Polyimide Synthesis and in Vivo Evaluation in Epileptic Mice.","authors":"Kshitij Kumar, Kaustubh Deshpande, Naveen Kalur, Garima Chauhan, Deepti Chugh, Subramaniam Ganesh, Arjun Ramakrishnan","doi":"10.1109/ACCESS.2026.3674701","DOIUrl":"10.1109/ACCESS.2026.3674701","url":null,"abstract":"<p><p>Thin-film polyimide neural probes have shown great promise in neuroscience but remain difficult to clinically translate due to the unavailability and lack of customizability of commercially available medical-grade polyamic acids. We present an open-source, end-to-end platform for synthesizing BPDA-pPDA-based polyimide from a custom polyamic acid and translating it into microfabricated thin-film neural interfaces. The approach combines accessible polymer chemistry with a streamlined MEMS-compatible fabrication process to produce flexible, biocompatible depth and surface electrode arrays with high thermal stability, chemical inertness, and low moisture uptake. Devices were validated through benchtop characterization, ISO 10993-11 systemic toxicity testing, and in vivo electrophysiology, both acute and semi-chronic, in wild-type and laforin knockout epileptic mice. The arrays reliably captured high-quality multi- and single-unit activity, as well as spontaneous epileptiform discharges, over implantation periods of up to 12 days. By demonstrating a customizable, end-to-end platform for synthesizing and fabricating thin-film polyimide neural electrodes, and by mimicking human neurosurgical workflows through depth, surface, and semi-chronic studies in mice, this work underscores the translational potential of polyimide-based neural microelectrodes and provides a practical pathway to accelerate clinical adoption.</p>","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"47115-47126"},"PeriodicalIF":3.6,"publicationDate":"2026-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC13089216/pdf/","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147722698","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A Procedural Architecture for Agent Trust and Credibility Verification in Distributed Ledger-Based Federated Learning 基于分布式账本的联邦学习中代理信任与可信度验证的过程化体系结构

IF 3.6 3区计算机科学

IEEE Access Pub Date : 2026-03-09 DOI: 10.1109/ACCESS.2026.3671712

Soonduck Yoo;Dae-Yeol Kim;Do-Yup Kim

{"title":"A Procedural Architecture for Agent Trust and Credibility Verification in Distributed Ledger-Based Federated Learning","authors":"Soonduck Yoo;Dae-Yeol Kim;Do-Yup Kim","doi":"10.1109/ACCESS.2026.3671712","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3671712","url":null,"abstract":"Agent-based federated learning (FL) enables multiple nodes to collaboratively train a global model without sharing raw data, thereby mitigating privacy and security concerns. However, FL inherently relies on the credibility of participating agents, making trust assurance critical in environments susceptible to malicious attacks and operational faults. This paper proposes a procedural architecture for evaluating and enhancing agent credibility in distributed ledger-based FL (DLFL) systems. The proposed architecture spans the agent life cycle, consisting of the pre-training, in-training, and post-training stages, and integrates three verification domains: data-based, behavior-based, and technology-based verification. In the pre-training stage, data-based verification assesses data integrity and quality through reference distribution analysis and entropy reduction to improve learning stability. During the in-training stage, behavior-based verification establishes a dual mechanism that monitors and analyzes agents’ learning processes and outcomes and detects anomalies in updates and performance. In the post-training stage, technology-based verification ensures record immutability and accountability through blockchain, cryptographic validation, and auditing mechanisms. By combining these multi-layered procedures, the proposed architecture enables systematic and continuous evaluation of agent credibility, fostering a trustworthy FL ecosystem and enabling future applications in autonomous agent collaboration and trust-oriented AI governance.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"38361-38374"},"PeriodicalIF":3.6,"publicationDate":"2026-03-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11424388","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147440592","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Cite and Seek: Automated Literary Reference Mining at Corpus Scale 引用和查找：语料库规模的自动文学参考挖掘

IF 3.6 3区计算机科学

IEEE Access Pub Date : 2026-03-09 DOI: 10.1109/ACCESS.2026.3672034

Shahar Golan;Chaya Liebeskind

{"title":"Cite and Seek: Automated Literary Reference Mining at Corpus Scale","authors":"Shahar Golan;Chaya Liebeskind","doi":"10.1109/ACCESS.2026.3672034","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3672034","url":null,"abstract":"Identifying intertextual references in literature is a complex task that requires both linguistic and contextual understanding. This study presents a methodology for extracting explicit intertextual references between books, focusing on explicit mentions of book titles, character names, and signature quotes. In contrast to scientific citation analysis, which relies on structured references, literary references are often implicit and require deeper semantic interpretation. We propose a multiphase approach that combines manual annotation, AI-assisted validation, and systematic disagreement analysis to construct a high-quality dataset of positive and negative examples from a corpus of approximately 30,000 public-domain books. We evaluate the performance of individual annotators and introduce an aggregate annotator that achieves an F1 score above 0.96 across all reference types. Our methodology supports a rigorous annotation process that reduces both false positives and false negatives. The resulting dataset offers a foundation for training and evaluating models capable of detecting literary connections, enabling deeper insights into intertextual influences in narrative literature.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"38375-38391"},"PeriodicalIF":3.6,"publicationDate":"2026-03-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11424554","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147440533","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

RSMA-Enabled RIS-Assisted Integrated Sensing and Communication for 6G: A Comprehensive Survey 支持rsma的ris辅助6G集成传感和通信：综合调查

IF 3.6 3区计算机科学

IEEE Access Pub Date : 2026-03-09 DOI: 10.1109/ACCESS.2026.3672355

Abdul-Manan Zakari Adams;Ahmed Abdelhadi

{"title":"RSMA-Enabled RIS-Assisted Integrated Sensing and Communication for 6G: A Comprehensive Survey","authors":"Abdul-Manan Zakari Adams;Ahmed Abdelhadi","doi":"10.1109/ACCESS.2026.3672355","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3672355","url":null,"abstract":"The evolution toward 6G wireless networks demands communication systems that are highly efficient, intelligent, and perceptive of their environment. Rate-splitting multiple access (RSMA), reconfigurable intelligent surface (RIS), and integrated sensing and communication (ISAC) have emerged as transformative technologies to meet these requirements. RSMA provides a new way to flexibly manage interference, RIS provides a controllable wireless environment, and ISAC improves spectral, energy, and hardware efficiencies. This survey presents a comprehensive study of RSMA-enabled RIS-assisted ISAC systems, covering the fundamental principles of RSMA and RIS, performance metrics, and optimization techniques. We also explore how their integration can markedly improve the performance of future wireless networks. The survey provides a comparative performance analysis with non-orthogonal multiple access (NOMA) and space division multiple access (SDMA) under various environmental setups and practical constraints. In particular, the paper studies 20 articles on RSMA-based RIS-assisted ISAC systems, where RSMA consistently demonstrates superior performance in terms of spectral efficiency, energy efficiency, and sensing accuracy compared to NOMA and SDMA. Additionally, we examine the role of artificial intelligence (AI) and machine learning (ML) in addressing the non-convex optimization challenges of RSMA-enabled RIS-assisted ISAC systems, emphasizing the potential of generative AI, deep reinforcement learning, and quantum machine learning. The survey outlines open research problems related to hardware impairments and security and identifies promising research directions, including integration with enabling 6G technologies. This survey serves as a comprehensive reference for researchers and professionals aiming to design intelligent, secure, and adaptive 6G wireless networks empowered by RSMA-enabled RIS-assisted ISAC systems.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"38392-38420"},"PeriodicalIF":3.6,"publicationDate":"2026-03-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11426895","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147440538","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Reinforcement Learning-Based Fuzzer for 5G RRC Security Evaluation 基于强化学习的5G RRC安全评估模糊器

IF 3.6 3区计算机科学

IEEE Access Pub Date : 2026-03-09 DOI: 10.1109/ACCESS.2026.3672003

Dhairya P. Parikh;Dimitri A. Dessources;Nishith D. Tripathi;Jeffrey H. Reed;Eric W. Burger

{"title":"Reinforcement Learning-Based Fuzzer for 5G RRC Security Evaluation","authors":"Dhairya P. Parikh;Dimitri A. Dessources;Nishith D. Tripathi;Jeffrey H. Reed;Eric W. Burger","doi":"10.1109/ACCESS.2026.3672003","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3672003","url":null,"abstract":"Open Radio Access Network (O-RAN) and modern Fifth Generation Mobile Networks (5G) Standalone (SA) deployments increase protocol complexity and broaden the attack surface of cellular infrastructure. This paper introduces a reinforcement-learning-based fuzz tester designed to evaluate the Radio Resource Control (RRC) layer in 5G SA networks. The fuzzer operates as a software-defined “false” User Equipment (UE) that attaches to the target network, intercepts and mutates uplink RRC messages, and injects malformed test cases targeting RRC handlers. The system integrates Reinforcement Learning (RL)- driven test-case generation with an automated execution pipeline for message injection and packet-capture analysis, allowing the agent to iteratively learn which mutations most effectively trigger anomalous behavior. Reinforcement feedback is computed from system metrics such as Central Processing Unit (CPU) utilization, thread count, and network Input/Output (I/O) to guide learning toward high-impact inputs. Experimental results demonstrate that the proposed fuzzer uncovers previously unseen protocol-handling anomalies, malformed-message behaviors, and resource-exhaustion conditions, including reproducible RRC/NGAP inconsistencies identified through a deterministic Proof-of-Concept (PoC) evaluation. The paper presents the overall architecture, reinforcement learning formulation, and evaluation results, highlighting how feedbackdriven adaptive fuzzing can prioritize high-impact mutations for stateful 5G RRC security assessment.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"38264-38274"},"PeriodicalIF":3.6,"publicationDate":"2026-03-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11424421","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147440530","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Named Entity Recognition With Clue-Word Tags From Patent Documents in Materials Science 基于材料科学专利文献线索词标签的命名实体识别

IF 3.6 3区计算机科学

IEEE Access Pub Date : 2026-03-09 DOI: 10.1109/ACCESS.2026.3671976

Toshihiko Sakai;Nobuhiko Chiwata;Tsunenori Mine

{"title":"Named Entity Recognition With Clue-Word Tags From Patent Documents in Materials Science","authors":"Toshihiko Sakai;Nobuhiko Chiwata;Tsunenori Mine","doi":"10.1109/ACCESS.2026.3671976","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3671976","url":null,"abstract":"In the field of materials science, it is important to extract expressions such as composition and ratios of materials from patent documents for the investigation of new material development. However, numerical values representing ratios can also appear in contexts unrelated to ratios, leading to ambiguity during extraction. Therefore, this study proposes a method to assign ‘clue-word tags’ to information called ‘clue words’, which co-occurs with numerical values representing ratios, to facilitate the extraction of ratios. Explicitly identifying clue words makes it easier to extract numerical values with meanings as ratios and compositions appearing near these ratio-related information. This concept can be generalized to the extraction of target words other than ratios. Experimental results on Japanese patent documents show that clue-word tags improved the extraction performance, specifically, for ‘fig_LL’ (lower limit of element content; outperforming by 0.0135 points). To validate the generalizability of the clue-word tag hypothesis, we created an English patent dataset of 10,166 sentences through semi-supervised learning with RoBERTa-large and conducted 5-fold cross-validation experiments using CRF. Results on the English dataset demonstrate that clue-word tags consistently improve performance across all feature patterns, with significant improvements in micro F1-scores (p= 0.0312) and particularly strong effects on ‘fig_LL’ recognition (+ 0.0212 to + 0.0344 improvement). Furthermore, applying regular expressions enhanced extraction for specific tags, and merging predictions from models trained with and without clue-word tags further improved overall performance. Additionally, the proposed method simultaneously discovers candidate clue words during the named entity recognition process.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"38332-38346"},"PeriodicalIF":3.6,"publicationDate":"2026-03-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11426775","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147440510","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Development of a Neural Network-Based Model to Generate an Absolute Luminance Map of an Interior Using a Camera Raw Image File 利用相机原始图像文件生成室内绝对亮度图的基于神经网络模型的开发

IF 3.6 3区计算机科学

IEEE Access Pub Date : 2026-03-09 DOI: 10.1109/ACCESS.2026.3672040

Vedavyasa Kamath;Ciji Pearl Kurian;K. R. Shailesh;U. Suprabha Padiyar

{"title":"Development of a Neural Network-Based Model to Generate an Absolute Luminance Map of an Interior Using a Camera Raw Image File","authors":"Vedavyasa Kamath;Ciji Pearl Kurian;K. R. Shailesh;U. Suprabha Padiyar","doi":"10.1109/ACCESS.2026.3672040","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3672040","url":null,"abstract":"This article proposes a novel machine learning-based technique for estimating absolute luminance maps of interior scenes illuminated by artificial lights. The method involves using a single raw image captured by a Digital Single-Lens Reflex (DSLR) camera, which is then processed through a neural network estimation model based on Bayesian regularization. The model is trained on a dataset of linear R, G, and B pixel values of ColorChecker Classic chart swatches and their corresponding actual luminance values measured by a Konica Minolta LS 150 luminance meter across a wide range of lighting levels. The article discusses the tools and techniques used to develop the model, as well as the image capturing and processing stages involved in generating 14-bit R, G, and B channel pixel values. These pixel values serve as inputs to the model, which predicts absolute luminance at the corresponding pixel position in the scene, thereby generating a luminance map. The proposed technique is cost-effective, easy to perform, and scientifically viable, with estimation errors of less than 10 percent on surfaces made of paint, paper, and cloth. The technique has potential applications in optimizing artificial lighting, daylight design, and enhancing visual comfort in various scenarios. The technique also offers advantages over existing high dynamic range imaging methods, which often require high lighting consistency during image capture and computationally intensive image processing. Overall, the proposed technique presents a promising new approach to accurate luminance mapping, with broad practical implications.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"38291-38305"},"PeriodicalIF":3.6,"publicationDate":"2026-03-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11424583","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147440512","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Sound Event Detection System With Frequency-Aware Enhancements and Semi-Supervised Learning 具有频率感知增强和半监督学习的声音事件检测系统

IF 3.6 3区计算机科学

IEEE Access Pub Date : 2026-03-06 DOI: 10.1109/ACCESS.2026.3671264

Narin Kim;Sumi Lee;Sojung Jang;Juhyun Lee;Il-Youp Kwak

{"title":"Sound Event Detection System With Frequency-Aware Enhancements and Semi-Supervised Learning","authors":"Narin Kim;Sumi Lee;Sojung Jang;Juhyun Lee;Il-Youp Kwak","doi":"10.1109/ACCESS.2026.3671264","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3671264","url":null,"abstract":"Sound Event Detection (SED) systems are essential for understanding and classifying the causes and temporal occurrences of sounds in diverse environments. This paper introduces a robust and efficient SED system that integrates a novel Frequency-aware Lightweight Convolutional Attention Module (FLCAM) and semi-supervised learning techniques to address key challenges in audio analysis. The FLCAM enhances 2D convolutional models by preserving critical frequency-domain features and adaptively assigning attention weights, enabling superior performance while maintaining computational efficiency. To fully leverage strongly labeled, weakly labeled, and unlabeled data, our system employs the Mean Teacher framework, which ensures consistency between predictions under different augmentations. Comprehensive experiments on the DESED and L3DAS22 datasets demonstrate the effectiveness of our approach, achieving improvements of approximately 9 percentage points in PSDS and 2 percentage points in F-score metrics, respectively. Despite utilizing significantly fewer parameters, the proposed SED system achieves performance comparable to state-of-the-art models, making it suitable for real-world applications, including resource-constrained environments.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"38347-38360"},"PeriodicalIF":3.6,"publicationDate":"2026-03-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11422831","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147440534","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

DFT-Assisted Multimodulus Blind Equalization for GFDM dft辅助的GFDM多模盲均衡

IF 3.6 3区计算机科学

IEEE Access Pub Date : 2026-03-06 DOI: 10.1109/ACCESS.2026.3671184

Effrina Yanti Hamid;Bella Wahmilyana Asril

{"title":"DFT-Assisted Multimodulus Blind Equalization for GFDM","authors":"Effrina Yanti Hamid;Bella Wahmilyana Asril","doi":"10.1109/ACCESS.2026.3671184","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3671184","url":null,"abstract":"As sixth-generation (6G) network development progresses, research continues to explore various technologies and new waveforms to meet the demands of future communications. Generalized frequency division multiplexing (GFDM) is considered as a promising waveform due to its adaptable structure, reduced latency, and efficient use of the spectrum. Nevertheless, nonorthogonal subcarriers of GFDM induce severe inter-symbol interference (ISI) and inter-carrier interference (ICI), leading to slow convergence and elevated symbol error rate (SER) floors. This study introduces a blind equalization strategy that integrates the multimodulus algorithm (MMA) with discrete Fourier transform (DFT)-based channel estimation. The simulation results show that a GFDM system employing the proposed MMA-DFT blind equalizer with <inline-formula> <tex-math>$K=128$ </tex-math></inline-formula> subcarriers and <inline-formula> <tex-math>$M=5$ </tex-math></inline-formula> subsymbols over the long-term evolution extended pedestrian A (LTE-EPA) channel closely approaches the theoretical bound within 20 GFDM-block iterations. Moreover, the mean square error (MSE) of MMA-DFT converges within approximately 15–20 GFDM blocks, enabling near-optimal detection performance with a limited number of adaptation iterations. Furthermore, as a result of the higher effective signal-to-noise ratio (SNR) enabled by the reduced cyclic prefix (CP) overhead of GFDM, the MMA-DFT equalizer applied to GFDM achieves a 0.89-dB SER gain compared to the same MMA-DFT equalizer applied to orthogonal frequency division multiplexing (OFDM) under identical parameter settings. A theoretical SER expression incorporating misadjustment, channel frequency response (CFR)-estimation error, and residual ICI is derived and shown to closely match the simulation. The proposed method reduces pilot overhead, improves spectral efficiency, and supports scalable massive multiple-input multiple-output (mMIMO) scenarios, contributing to energy-efficient future networks.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"38275-38290"},"PeriodicalIF":3.6,"publicationDate":"2026-03-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11422838","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147440485","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

A Regional Benchmark for Deep Learning-Based Hourly Precipitation Nowcasting in Latin America 拉丁美洲基于深度学习的逐时降水临近预报的区域基准

IF 3.6 3区计算机科学

IEEE Access Pub Date : 2026-03-05 DOI: 10.1109/ACCESS.2026.3670767

Adriano P. Almeida;Henrique M. J. Barbosa;Sâmia R. Garcia;David J. Gagne;Kanghui Zhou;Takuji Kubota;Tomoo Ushio;Shigenori Otsuka;Simon Pfreundschuh;Alan J. P. Calheiros

{"title":"A Regional Benchmark for Deep Learning-Based Hourly Precipitation Nowcasting in Latin America","authors":"Adriano P. Almeida;Henrique M. J. Barbosa;Sâmia R. Garcia;David J. Gagne;Kanghui Zhou;Takuji Kubota;Tomoo Ushio;Shigenori Otsuka;Simon Pfreundschuh;Alan J. P. Calheiros","doi":"10.1109/ACCESS.2026.3670767","DOIUrl":"https://doi.org/10.1109/ACCESS.2026.3670767","url":null,"abstract":"Accurate short-term precipitation forecasting is critical for Latin America, but the region lacks a standardized framework to evaluate data-driven approaches due to the sparse coverage in the ground. This study introduces the Artificial Intelligence for Nowcasting Pilot Project (AINPP) Precipitation Benchmark (AINPP-PB-LATAM), providing curated datasets and a scalable, optimized training pipeline designed to accelerate deep learning development in high-performance computing environments. Beyond establishing a baseline using seven years of satellite-based data (2018–2024), the framework reduces engineering barriers, enabling researchers to focus on fine-tuning strategies to extract the full predictive capacity of models for regional specificities. As a demonstration use case, we trained and evaluated five deep learning architectures, AFNO, Inception-V4, ResNet-50, U-Net, and Xception, comparing them against Lagrangian Persistence and the operational AI Nowcast. The results reveal critical trade-offs: spectral methods like AFNO excel in continuous skill by capturing large-scale dependencies, while convolutional architectures offer robust categorical performance. However, pixel-wise optimization challenges persist, with systematic under-prediction of heavy rainfall. By providing open-access code and optimized baseline implementations for distributed computing, AINPP-PB-LATAM establishes a scalable foundation for collaborative research, facilitating the advancement of operational AI-based nowcasting and transferability assessments in data-scarce regions.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"14 ","pages":"38306-38331"},"PeriodicalIF":3.6,"publicationDate":"2026-03-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11421841","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"147440532","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":3,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

IEEE Access最新文献