While Convolutional Neural Networks (CNNs) excel at learning complex latent-space representations, their over-parameterization can lead to overfitting and reduced performance, particularly with limited data. This, alongside their high computational and memory demands, limits the applicability of CNNs for edge deployment and applications where computational resources are constrained. Low-rank matrix approximation has emerged as a promising approach to reduce CNN parameters, but existing methods often require pre-determined ranks or involve complex post-training adjustments, leading to challenges in rank selection, performance loss, and limited practicality in resource-constrained environments. This underscores the need for an adaptive compression method that integrates into the training process, dynamically adjusting model complexity based on data and task requirements. To address this, we propose an efficient training method for CNN compression via dynamic parameter rank pruning. Our approach integrates efficient matrix factorization and novel regularization techniques, forming a robust framework for dynamic rank pruning and model compression. By using Singular Value Decomposition (SVD) to model low-rank convolutional filters and dense weight matrices, and training the SVD factors with back-propagation in an end-to-end manner, we achieve model compression. We evaluate our method on modern CNNs, including ResNet-18, ResNet-20, and ResNet-32, using datasets like CIFAR-10, CIFAR-100, and ImageNet (2012). Our experiments demonstrate that the proposed method can reduce model parameters by up to 50% and improve classification accuracy by up to 2% over baseline models, making CNNs more feasible for practical applications.
{"title":"Convolutional Neural Network Compression via Dynamic Parameter Rank Pruning","authors":"Manish Sharma;Jamison Heard;Eli Saber;Panagiotis Markopoulos","doi":"10.1109/ACCESS.2025.3533419","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3533419","url":null,"abstract":"While Convolutional Neural Networks (CNNs) excel at learning complex latent-space representations, their over-parameterization can lead to overfitting and reduced performance, particularly with limited data. This, alongside their high computational and memory demands, limits the applicability of CNNs for edge deployment and applications where computational resources are constrained. Low-rank matrix approximation has emerged as a promising approach to reduce CNN parameters, but existing methods often require pre-determined ranks or involve complex post-training adjustments, leading to challenges in rank selection, performance loss, and limited practicality in resource-constrained environments. This underscores the need for an adaptive compression method that integrates into the training process, dynamically adjusting model complexity based on data and task requirements. To address this, we propose an efficient training method for CNN compression via dynamic parameter rank pruning. Our approach integrates efficient matrix factorization and novel regularization techniques, forming a robust framework for dynamic rank pruning and model compression. By using Singular Value Decomposition (SVD) to model low-rank convolutional filters and dense weight matrices, and training the SVD factors with back-propagation in an end-to-end manner, we achieve model compression. We evaluate our method on modern CNNs, including ResNet-18, ResNet-20, and ResNet-32, using datasets like CIFAR-10, CIFAR-100, and ImageNet (2012). Our experiments demonstrate that the proposed method can reduce model parameters by up to 50% and improve classification accuracy by up to 2% over baseline models, making CNNs more feasible for practical applications.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"18441-18456"},"PeriodicalIF":3.4,"publicationDate":"2025-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10851278","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143105844","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}
Pub Date : 2025-01-23DOI: 10.1109/ACCESS.2025.3533099
Chunxiao Wu
Rural energy has the characteristics of diversity, dispersion, and regionalism. The construction and operation of its comprehensive energy system are essential for promoting the development of a low-carbon economy in rural areas and reducing carbon emissions. In response, the structure of the rural integrated energy system is initially analyzed and the equipment is planned. Subsequently, the carbon emission in each link of the power generation system is estimated by using the life cycle assessment method. Based on the evaluation results, a suitable low-carbon economy optimal scheduling model is established. The generator set and energy storage equipment in the integrated energy system are designed to enhance the efficiency of energy utilization. Then, the life cycle assessment method is used to calculate the carbon emissions of each stage. Next, the optimal scheduling model is established by taking the minimum carbon emissions and system operating costs as the objective function. In addition, the study introduced a water method and a kinetic accelerator to capture carbon dioxide and feed it into a high-pressure pump for storage. This will enable carbon capture and net zero emissions. The findings revealed that Model 1 exhibited minimal carbon emissions, amounting to 48,530.5 kg, and a comparatively low total operating cost of 30,625.48. After being applied to practical systems, the carbon emissions have been significantly reduced. The research model can effectively guide the low-carbon transformation of rural integrated energy systems and provide important technical support for achieving sustainable development goals in rural areas.
{"title":"Carbon Emission Evaluation and Low Carbon Economy Optimization Scheduling of Rural Integrated Energy System Based on LCA Method","authors":"Chunxiao Wu","doi":"10.1109/ACCESS.2025.3533099","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3533099","url":null,"abstract":"Rural energy has the characteristics of diversity, dispersion, and regionalism. The construction and operation of its comprehensive energy system are essential for promoting the development of a low-carbon economy in rural areas and reducing carbon emissions. In response, the structure of the rural integrated energy system is initially analyzed and the equipment is planned. Subsequently, the carbon emission in each link of the power generation system is estimated by using the life cycle assessment method. Based on the evaluation results, a suitable low-carbon economy optimal scheduling model is established. The generator set and energy storage equipment in the integrated energy system are designed to enhance the efficiency of energy utilization. Then, the life cycle assessment method is used to calculate the carbon emissions of each stage. Next, the optimal scheduling model is established by taking the minimum carbon emissions and system operating costs as the objective function. In addition, the study introduced a water method and a kinetic accelerator to capture carbon dioxide and feed it into a high-pressure pump for storage. This will enable carbon capture and net zero emissions. The findings revealed that Model 1 exhibited minimal carbon emissions, amounting to 48,530.5 kg, and a comparatively low total operating cost of 30,625.48. After being applied to practical systems, the carbon emissions have been significantly reduced. The research model can effectively guide the low-carbon transformation of rural integrated energy systems and provide important technical support for achieving sustainable development goals in rural areas.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"17182-17194"},"PeriodicalIF":3.4,"publicationDate":"2025-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10851269","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143105482","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}
Pub Date : 2025-01-23DOI: 10.1109/ACCESS.2025.3533137
Satyaprakash Rout;Satyajit Das
The accuracy of model-based State of Charge (SOC) estimators often degrades due to parametric uncertainty, measurement errors, and variations in operating temperature. Many Kalman filter-driven SOC estimators in the literature overlook these uncertainties, leading to imprecise SOC estimation. To address these challenges, this study proposes a joint forgetting factor-based adaptive extended Kalman filter (JFFAEKF). The JFFAEKF approach evaluates uncertainties in the battery model and incorporates them into the SOC estimation process under dynamic operating conditions. By augmenting both the SOC and battery model parameters into a single state vector, the estimator concurrently updates these variables. An adaptive correction mechanism for process and measurement noise covariance matrices is introduced, leveraging the innovation and residual errors of estimated terminal voltage. These covariance updates enable the computation of an appropriate filter gain to mitigate the adverse effects of model and measurement uncertainties. Additionally, a forgetting factor is integrated into the design to enhance computational efficiency and convergence rate. The practical applicability of the proposed JFFAEKF is validated using real-world current profiles from the LA92, UDDS, and US06 drive cycles at various operating temperatures. The accuracy of the SOC estimation is demonstrated by comparing the root mean square error ($E_{RMS}$ ) and maximum absolute error ($Max_{AE}$ ) with other Kalman filter-based estimators. Furthermore, the estimator’s robustness is tested under adverse conditions, including offset current, sensor bias voltage, and parametric uncertainties in the battery model and state estimator. Results from diverse dynamic operating conditions confirm the superior performance of the JFFAEKF in SOC estimation compared to existing methods.
{"title":"A Joint Forgetting Factor-Based Adaptive Extended Kalman Filtering Approach to Predict the State-of-Charge and Model Parameter of Lithium-Ion Battery","authors":"Satyaprakash Rout;Satyajit Das","doi":"10.1109/ACCESS.2025.3533137","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3533137","url":null,"abstract":"The accuracy of model-based State of Charge (SOC) estimators often degrades due to parametric uncertainty, measurement errors, and variations in operating temperature. Many Kalman filter-driven SOC estimators in the literature overlook these uncertainties, leading to imprecise SOC estimation. To address these challenges, this study proposes a joint forgetting factor-based adaptive extended Kalman filter (JFFAEKF). The JFFAEKF approach evaluates uncertainties in the battery model and incorporates them into the SOC estimation process under dynamic operating conditions. By augmenting both the SOC and battery model parameters into a single state vector, the estimator concurrently updates these variables. An adaptive correction mechanism for process and measurement noise covariance matrices is introduced, leveraging the innovation and residual errors of estimated terminal voltage. These covariance updates enable the computation of an appropriate filter gain to mitigate the adverse effects of model and measurement uncertainties. Additionally, a forgetting factor is integrated into the design to enhance computational efficiency and convergence rate. The practical applicability of the proposed JFFAEKF is validated using real-world current profiles from the LA92, UDDS, and US06 drive cycles at various operating temperatures. The accuracy of the SOC estimation is demonstrated by comparing the root mean square error (<inline-formula> <tex-math>$E_{RMS}$ </tex-math></inline-formula>) and maximum absolute error (<inline-formula> <tex-math>$Max_{AE}$ </tex-math></inline-formula>) with other Kalman filter-based estimators. Furthermore, the estimator’s robustness is tested under adverse conditions, including offset current, sensor bias voltage, and parametric uncertainties in the battery model and state estimator. Results from diverse dynamic operating conditions confirm the superior performance of the JFFAEKF in SOC estimation compared to existing methods.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"16770-16786"},"PeriodicalIF":3.4,"publicationDate":"2025-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10851262","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143106002","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}
Advanced Driver Assistance System (ADAS) has become an essential feature in vehicles, and it is leading to the evolution of autonomous vehicles. But the technologies to implement ADAS suffer from certain inherent limitations, such as latency rate, computational speed, accuracy of the algorithm, security, and privacy, which are also the important factors for realizing full autonomous vehicles. With respect to these hindrances, an in-depth analysis of the existing research has shown that quantum machine learning (QML) can hold a powerful and alternate solution for the development of autonomous vehicles. The perks of quantum computation (QC) over classical systems are apparent with respect to security, privacy, and an exponentially high computation rate. The current review study underlines the benefits of quantum computation and asks for more QML research to improve real-time decision-making in autonomous vehicles, ultimately improving their safety and efficiency. The promise of quantum computing to handle the massive data and computational complexity that classical methods struggle with necessitates new studies in quantum machine learning (QML) for autonomous vehicles.
{"title":"Quantum Computing for Advanced Driver Assistance Systems and Autonomous Vehicles: A Review","authors":"Avantika Rattan;Abhishek Rudra Pal;Muralimohan Gurusamy","doi":"10.1109/ACCESS.2025.3532958","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3532958","url":null,"abstract":"Advanced Driver Assistance System (ADAS) has become an essential feature in vehicles, and it is leading to the evolution of autonomous vehicles. But the technologies to implement ADAS suffer from certain inherent limitations, such as latency rate, computational speed, accuracy of the algorithm, security, and privacy, which are also the important factors for realizing full autonomous vehicles. With respect to these hindrances, an in-depth analysis of the existing research has shown that quantum machine learning (QML) can hold a powerful and alternate solution for the development of autonomous vehicles. The perks of quantum computation (QC) over classical systems are apparent with respect to security, privacy, and an exponentially high computation rate. The current review study underlines the benefits of quantum computation and asks for more QML research to improve real-time decision-making in autonomous vehicles, ultimately improving their safety and efficiency. The promise of quantum computing to handle the massive data and computational complexity that classical methods struggle with necessitates new studies in quantum machine learning (QML) for autonomous vehicles.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"17554-17582"},"PeriodicalIF":3.4,"publicationDate":"2025-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10850907","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143106017","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}
Pub Date : 2025-01-23DOI: 10.1109/ACCESS.2025.3533217
Marco Siino;Mariana Falco;Daniele Croce;Paolo Rosso
Artificial Intelligence (AI) is reshaping the legal landscape, with software tools now impacting various aspects of legal work. The intersection of Natural Language Processing (NLP) and law holds potential to transform how legal professionals, including lawyers and judges, operate, resolve disputes, and retrieve case information to formulate their decisions. To identify the current state of the applications of Transformers (also known as Large Language Models or LLMs) in the legal domain, we analysed the existing literature from 2017 to 2023 through a database search and snowballing method. From 61 selected publications, we identified key application categories such as legal document analysis, case prediction, and contract review, along with their main characteristics. We observed a discernible upsurge in the volume of scholarly publications, a diversification of tasks undertaken (e.g., legal research, contract analysis, and regulatory compliance), and an increased range of languages considered. There has been a notable enhancement in the methodological sophistication employed by researchers in practical applications. The performance of models grounded in the Generative Pre-trained Transformer (GPT) architecture has consistently improved across various legal domains, including contract review, legal document summarization, and case outcome prediction. This paper makes several significant contributions to the field. Firstly, it identifies emerging trends in the application of LLMs within the legal domain, highlighting the growing interest and investment in this area. Secondly, it pinpoints methodological gaps in current research, suggesting areas where further development and refinement are needed. Lastly, it discusses the broader implications of these advancements for real-world legal tasks, offering insights into how LLM-based AI can enhance legal practice while addressing the associated challenges.
{"title":"Exploring LLMs Applications in Law: A Literature Review on Current Legal NLP Approaches","authors":"Marco Siino;Mariana Falco;Daniele Croce;Paolo Rosso","doi":"10.1109/ACCESS.2025.3533217","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3533217","url":null,"abstract":"Artificial Intelligence (AI) is reshaping the legal landscape, with software tools now impacting various aspects of legal work. The intersection of Natural Language Processing (NLP) and law holds potential to transform how legal professionals, including lawyers and judges, operate, resolve disputes, and retrieve case information to formulate their decisions. To identify the current state of the applications of Transformers (also known as Large Language Models or LLMs) in the legal domain, we analysed the existing literature from 2017 to 2023 through a database search and snowballing method. From 61 selected publications, we identified key application categories such as legal document analysis, case prediction, and contract review, along with their main characteristics. We observed a discernible upsurge in the volume of scholarly publications, a diversification of tasks undertaken (e.g., legal research, contract analysis, and regulatory compliance), and an increased range of languages considered. There has been a notable enhancement in the methodological sophistication employed by researchers in practical applications. The performance of models grounded in the Generative Pre-trained Transformer (GPT) architecture has consistently improved across various legal domains, including contract review, legal document summarization, and case outcome prediction. This paper makes several significant contributions to the field. Firstly, it identifies emerging trends in the application of LLMs within the legal domain, highlighting the growing interest and investment in this area. Secondly, it pinpoints methodological gaps in current research, suggesting areas where further development and refinement are needed. Lastly, it discusses the broader implications of these advancements for real-world legal tasks, offering insights into how LLM-based AI can enhance legal practice while addressing the associated challenges.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"18253-18276"},"PeriodicalIF":3.4,"publicationDate":"2025-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10850911","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143105505","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}
Pub Date : 2025-01-23DOI: 10.1109/ACCESS.2025.3532949
Rui Xue
The quality assessment of physical education programs in higher education is essential for fostering student development and institutional excellence. However, several uncertain and ambiguous factors cause the complexity of the quality assessment of physical education. This study aims to evaluate the quality of physical education, considering the uncertain factors using a well-known framework known as picture fuzzy set (PFS). This study introduces an innovative decision-making model by integrating the Multi-Objective Optimization by Ratio Analysis plus Full Multiplicative Form (MULTIMOORA) method and the Muirhead Mean (MM) operator. The PFS enables the model of the expert’s opinion using membership degree (MD), degrees of neutral membership (DONM), and non-membership degree (ND) and consequently effectively addresses uncertainty and ambiguity in multi-criteria decision-making problems. The MM operator enhances the accuracy by capturing interdependencies among evaluation criteria, ensuring more precise and comprehensive analyses. The MULTIMOORA method ensures robust analysis in the developed decision model for assessing the physical education quality because of the components of the Ratio System (RS), Reference Point Approach (RPA), and Full Multiplicative Form (FMF). The practical implications of this work are significant, as it equips stakeholders with actionable insights for curriculum development, resource optimization, and policy-making in physical education. A numerical example demonstrates the method’s utility in real-world scenarios, showcasing its effectiveness in addressing challenges inherent in higher education quality assessments. This study advances decision science by providing a scientifically rigorous and practically impactful tool for evaluating and improving physical education programs.
{"title":"Evaluating Physical Education Quality in Higher Education Using a Picture Fuzzy Decision Framework With Muirhead Mean Operator and MULTIMOORA Method","authors":"Rui Xue","doi":"10.1109/ACCESS.2025.3532949","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3532949","url":null,"abstract":"The quality assessment of physical education programs in higher education is essential for fostering student development and institutional excellence. However, several uncertain and ambiguous factors cause the complexity of the quality assessment of physical education. This study aims to evaluate the quality of physical education, considering the uncertain factors using a well-known framework known as picture fuzzy set (PFS). This study introduces an innovative decision-making model by integrating the Multi-Objective Optimization by Ratio Analysis plus Full Multiplicative Form (MULTIMOORA) method and the Muirhead Mean (MM) operator. The PFS enables the model of the expert’s opinion using membership degree (MD), degrees of neutral membership (DONM), and non-membership degree (ND) and consequently effectively addresses uncertainty and ambiguity in multi-criteria decision-making problems. The MM operator enhances the accuracy by capturing interdependencies among evaluation criteria, ensuring more precise and comprehensive analyses. The MULTIMOORA method ensures robust analysis in the developed decision model for assessing the physical education quality because of the components of the Ratio System (RS), Reference Point Approach (RPA), and Full Multiplicative Form (FMF). The practical implications of this work are significant, as it equips stakeholders with actionable insights for curriculum development, resource optimization, and policy-making in physical education. A numerical example demonstrates the method’s utility in real-world scenarios, showcasing its effectiveness in addressing challenges inherent in higher education quality assessments. This study advances decision science by providing a scientifically rigorous and practically impactful tool for evaluating and improving physical education programs.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"18277-18293"},"PeriodicalIF":3.4,"publicationDate":"2025-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10850913","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143105444","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}
Pub Date : 2025-01-23DOI: 10.1109/ACCESS.2025.3533023
Kyung-Duk Jang;Jae-Hyun Lee
A frequency-tunable tri-band metamaterial-inspired antenna is proposed. The antenna consists of patches, interdigital capacitors, and meander lines inductors, and loaded with a split ring resonator. The metamaterial-inspired antenna can provide three band operations at 1.575/2.1/2.6 GHz, which cover the GPS band1, LTE band1, and LTE band7 applications. The antenna has three different current flows, which create different resonances. The proposed antenna can adjust each resonant frequency independently by adjusting the antenna parameters, the length of SRR, meander lines, and ground branch line, which can affect each current path. The prototype antenna has been designed and fabricated using the FR4 substrate with a thickness of 1.6 mm and a dielectric constant of 4.3. The measured return loss and far-field radiation patterns of the fabricated antenna have good agreement with the simulated results. The fabricated antenna shows well-matched impedance characteristics with a reasonable bandwidth and has an omnidirectional radiation pattern at each operating band. The proposed antenna has the advantages of multi-band, frequency-tunable characteristics, and easy fabrication, and can be widely used in various wireless mobile communication systems.
{"title":"Frequency-Tunable Tri-Band Metamaterial-Inspired Antenna","authors":"Kyung-Duk Jang;Jae-Hyun Lee","doi":"10.1109/ACCESS.2025.3533023","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3533023","url":null,"abstract":"A frequency-tunable tri-band metamaterial-inspired antenna is proposed. The antenna consists of patches, interdigital capacitors, and meander lines inductors, and loaded with a split ring resonator. The metamaterial-inspired antenna can provide three band operations at 1.575/2.1/2.6 GHz, which cover the GPS band1, LTE band1, and LTE band7 applications. The antenna has three different current flows, which create different resonances. The proposed antenna can adjust each resonant frequency independently by adjusting the antenna parameters, the length of SRR, meander lines, and ground branch line, which can affect each current path. The prototype antenna has been designed and fabricated using the FR4 substrate with a thickness of 1.6 mm and a dielectric constant of 4.3. The measured return loss and far-field radiation patterns of the fabricated antenna have good agreement with the simulated results. The fabricated antenna shows well-matched impedance characteristics with a reasonable bandwidth and has an omnidirectional radiation pattern at each operating band. The proposed antenna has the advantages of multi-band, frequency-tunable characteristics, and easy fabrication, and can be widely used in various wireless mobile communication systems.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"21210-21215"},"PeriodicalIF":3.4,"publicationDate":"2025-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10851270","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143105564","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}
Pub Date : 2025-01-23DOI: 10.1109/ACCESS.2025.3532995
Tatsuya Tojima;Mitsuo Yoshida
Art has become an important new investment vehicle. Thus, interest is growing in art price prediction as a tool for assessing the returns and risks of art investments. Both traditional statistical methods and machine learning methods have been used to predict art prices. However, both methods incur substantial human costs for data preprocessing for the construction of prediction models, necessitating a reduction in the workload. In this study, we propose the zero-shot classification method to perform automatic annotation in data processing for art price prediction by leveraging large language models (LLMs). The proposed method can perform annotation without new training data. Thus, it minimizes human costs. Our experiments demonstrated that the 4-bit quantized Llama-3 70B model, which can run on a local server, achieved the most accurate (over 0.9) automatic annotation of different art forms using LLMs, performing slightly better than the GPT-4o model from OpenAI. These results are practical for data preprocessing and comparable with the results of previous machine learning methods.
{"title":"Zero-Shot Classification of Art With Large Language Models","authors":"Tatsuya Tojima;Mitsuo Yoshida","doi":"10.1109/ACCESS.2025.3532995","DOIUrl":"https://doi.org/10.1109/ACCESS.2025.3532995","url":null,"abstract":"Art has become an important new investment vehicle. Thus, interest is growing in art price prediction as a tool for assessing the returns and risks of art investments. Both traditional statistical methods and machine learning methods have been used to predict art prices. However, both methods incur substantial human costs for data preprocessing for the construction of prediction models, necessitating a reduction in the workload. In this study, we propose the zero-shot classification method to perform automatic annotation in data processing for art price prediction by leveraging large language models (LLMs). The proposed method can perform annotation without new training data. Thus, it minimizes human costs. Our experiments demonstrated that the 4-bit quantized Llama-3 70B model, which can run on a local server, achieved the most accurate (over 0.9) automatic annotation of different art forms using LLMs, performing slightly better than the GPT-4o model from OpenAI. These results are practical for data preprocessing and comparable with the results of previous machine learning methods.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"17426-17439"},"PeriodicalIF":3.4,"publicationDate":"2025-01-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10851281","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143105756","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}
Pub Date : 2025-01-22DOI: 10.1109/ACCESS.2024.3525276
Provides society information that may include news, reviews or technical notes that should be of interest to practitioners and researchers.
提供社会信息，可能包括新闻，评论或技术笔记，从业者和研究人员应该感兴趣。
{"title":"IEEE Access™ Editorial Board","authors":"","doi":"10.1109/ACCESS.2024.3525276","DOIUrl":"https://doi.org/10.1109/ACCESS.2024.3525276","url":null,"abstract":"Provides society information that may include news, reviews or technical notes that should be of interest to practitioners and researchers.","PeriodicalId":13079,"journal":{"name":"IEEE Access","volume":"13 ","pages":"i-x"},"PeriodicalIF":3.4,"publicationDate":"2025-01-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10849622","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"142993421","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}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：