Pub Date : 2025-11-01DOI: 10.1016/j.jiixd.2025.01.001
Xuan Wang , Danyang Yu , Yi Liu
In this paper, we propose an advanced multirelay-assisted hybrid (M-AH) simultaneous wireless information and power transfer (SWIPT) scheme to enhance the capacity in wireless communication systems. With the proposed scheme, the harvested energy at the relays within the same cluster can be utilized to improve the service quality of the optimal relay. Notably, the optimal relay is determined through an opportunistic relay selection approach. Moreover, we introduce a four-phase transmission strategy and develop an iterative optimization algorithm to maximize the system capacity (SC) while considering time slot and power constraints. The simulation results demonstrate that our proposed scheme outperforms existing schemes.
{"title":"Capacity enhancement in multirelay-assisted hybrid SWIPT wireless communications","authors":"Xuan Wang , Danyang Yu , Yi Liu","doi":"10.1016/j.jiixd.2025.01.001","DOIUrl":"10.1016/j.jiixd.2025.01.001","url":null,"abstract":"<div><div>In this paper, we propose an advanced multirelay-assisted hybrid (M-AH) simultaneous wireless information and power transfer (SWIPT) scheme to enhance the capacity in wireless communication systems. With the proposed scheme, the harvested energy at the relays within the same cluster can be utilized to improve the service quality of the optimal relay. Notably, the optimal relay is determined through an opportunistic relay selection approach. Moreover, we introduce a four-phase transmission strategy and develop an iterative optimization algorithm to maximize the system capacity (SC) while considering time slot and power constraints. The simulation results demonstrate that our proposed scheme outperforms existing schemes.</div></div>","PeriodicalId":100790,"journal":{"name":"Journal of Information and Intelligence","volume":"3 6","pages":"Pages 504-514"},"PeriodicalIF":0.0,"publicationDate":"2025-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145665461","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-01DOI: 10.1016/j.jiixd.2025.09.001
Mohammad Partohaghighi , Roummel F. Marcia , Bruce J. West , YangQuan Chen
This paper presents a unified framework for roughness-informed machine learning, dividing roughness into four categories: statistical, geometric, manifold, and topological. Statistical roughness, analyzed with tools like WeightWatcher, utilizes heavy-tailed weight distributions. Geometric roughness, measured by a novel roughness index, quantifies oscillatory patterns in loss landscapes. Manifold roughness, captured by the two-scale effective dimension, integrates local geometry (via fisher information matrix) with global parameter space complexity. Topological roughness, derived from persistence diagrams, evaluates structural complexity of learned functions. Experiments on MNIST, CIFAR-10, CIFAR-100, a damped harmonic oscillator, fractional order ODE, and wave equation demonstrate the framework's effectiveness: statistical roughness enhances federated learning convergence, geometric roughness improves training stability, manifold roughness optimizes generalization through noise injection, and topological roughness ensures smoother, physically accurate solutions. The framework advances model design, optimization, and generalization, with links to fractal and fractional calculus.
{"title":"Roughness-informed machine learning – A call for fractal and fractional calculi","authors":"Mohammad Partohaghighi , Roummel F. Marcia , Bruce J. West , YangQuan Chen","doi":"10.1016/j.jiixd.2025.09.001","DOIUrl":"10.1016/j.jiixd.2025.09.001","url":null,"abstract":"<div><div>This paper presents a unified framework for roughness-informed machine learning, dividing roughness into four categories: statistical, geometric, manifold, and topological. Statistical roughness, analyzed with tools like WeightWatcher, utilizes heavy-tailed weight distributions. Geometric roughness, measured by a novel roughness index, quantifies oscillatory patterns in loss landscapes. Manifold roughness, captured by the two-scale effective dimension, integrates local geometry (via fisher information matrix) with global parameter space complexity. Topological roughness, derived from persistence diagrams, evaluates structural complexity of learned functions. Experiments on MNIST, CIFAR-10, CIFAR-100, a damped harmonic oscillator, fractional order ODE, and wave equation demonstrate the framework's effectiveness: statistical roughness enhances federated learning convergence, geometric roughness improves training stability, manifold roughness optimizes generalization through noise injection, and topological roughness ensures smoother, physically accurate solutions. The framework advances model design, optimization, and generalization, with links to fractal and fractional calculus.</div></div>","PeriodicalId":100790,"journal":{"name":"Journal of Information and Intelligence","volume":"3 6","pages":"Pages 463-480"},"PeriodicalIF":0.0,"publicationDate":"2025-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145665464","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-01DOI: 10.1016/j.jiixd.2025.08.004
Gevira Omondi , Thomas O. Olwal
Multiple input multiple output (MIMO) communication systems have emerged as a key technol-ogy to enhance spectral efficiency and reliability in wireless communications. In recent years, deep neural network (DNN)-based approaches have shown promise in addressing the challenges of MIMO signal detection. Among these approaches, the Transformer architecture, known for its effectiveness in capturing long-range dependencies in sequential data, has gained significant attention. Therefore, this paper proposes a revolutionary DNN-based MIMO signal detection scheme using the Transformer-based architecture. This novel scheme leverages the multi-head self-attention mechanism inherent in Transformer architectures, which enables the model to capture both spatial and temporal dependencies in MIMO channels, thereby improving symbol detection accuracy and robustness under varying channel conditions. The proposed scheme's bit error rate (BER) performance is compared with traditional methods through simulations. The results show that the proposed method achieves a signal-to-noise ratio (SNR) gain of nearly 1.5 dB against the traditional detection methods, with the optimal maximum likelihood detector (MLD) only outperforming it by < 0.5 dB.
{"title":"A DNN-based MIMO signal detector using transformer architecture for next-generation wireless networks","authors":"Gevira Omondi , Thomas O. Olwal","doi":"10.1016/j.jiixd.2025.08.004","DOIUrl":"10.1016/j.jiixd.2025.08.004","url":null,"abstract":"<div><div>Multiple input multiple output (MIMO) communication systems have emerged as a key technol-ogy to enhance spectral efficiency and reliability in wireless communications. In recent years, deep neural network (DNN)-based approaches have shown promise in addressing the challenges of MIMO signal detection. Among these approaches, the Transformer architecture, known for its effectiveness in capturing long-range dependencies in sequential data, has gained significant attention. Therefore, this paper proposes a revolutionary DNN-based MIMO signal detection scheme using the Transformer-based architecture. This novel scheme leverages the multi-head self-attention mechanism inherent in Transformer architectures, which enables the model to capture both spatial and temporal dependencies in MIMO channels, thereby improving symbol detection accuracy and robustness under varying channel conditions. The proposed scheme's bit error rate (BER) performance is compared with traditional methods through simulations. The results show that the proposed method achieves a signal-to-noise ratio (SNR) gain of nearly 1.5 dB against the traditional detection methods, with the optimal maximum likelihood detector (MLD) only outperforming it by < 0.5 dB.</div></div>","PeriodicalId":100790,"journal":{"name":"Journal of Information and Intelligence","volume":"3 6","pages":"Pages 526-546"},"PeriodicalIF":0.0,"publicationDate":"2025-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145665463","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-01DOI: 10.1016/j.jiixd.2025.06.001
Helei Qiu , Biao Hou
In weakly supervised video anomaly detection (WSVAD) tasks, the temporal relationships of video are crucial for modeling event patterns. Transformer is a commonly used method for modeling temporal relationships. However, due to the large amount of redundancy in videos and the quadratic complexity of the Transformer, this method cannot effectively model long-range information. In addition, most WSVAD methods select key snippets based on predicted scores to represent event patterns, but this paradigm is susceptible to noise interference. To address the above issues, a novel temporal context and representative feature learning (TCRFL) method for WSVAD is proposed. Specifically, a temporal context learning (TCL) module is proposed to utilize both Mamba with linear complexity and Transformer to capture short-range and long-range dependencies of events. In addition, a representative feature learning (RFL) module is proposed to mine representative snippets to capture important information about events, further spreading it to video features to enhance the influence of representative features. The RFL module not only suppresses noise interference but also guides the model to select key snippets more accurately. The experimental results on UCF-Crime, XD-Violence, and ShanghaiTech datasets demonstrate the effectiveness and superiority of our method.
{"title":"Temporal context and representative feature learning for weakly supervised video anomaly detection","authors":"Helei Qiu , Biao Hou","doi":"10.1016/j.jiixd.2025.06.001","DOIUrl":"10.1016/j.jiixd.2025.06.001","url":null,"abstract":"<div><div>In weakly supervised video anomaly detection (WSVAD) tasks, the temporal relationships of video are crucial for modeling event patterns. Transformer is a commonly used method for modeling temporal relationships. However, due to the large amount of redundancy in videos and the quadratic complexity of the Transformer, this method cannot effectively model long-range information. In addition, most WSVAD methods select key snippets based on predicted scores to represent event patterns, but this paradigm is susceptible to noise interference. To address the above issues, a novel temporal context and representative feature learning (TCRFL) method for WSVAD is proposed. Specifically, a temporal context learning (TCL) module is proposed to utilize both Mamba with linear complexity and Transformer to capture short-range and long-range dependencies of events. In addition, a representative feature learning (RFL) module is proposed to mine representative snippets to capture important information about events, further spreading it to video features to enhance the influence of representative features. The RFL module not only suppresses noise interference but also guides the model to select key snippets more accurately. The experimental results on UCF-Crime, XD-Violence, and ShanghaiTech datasets demonstrate the effectiveness and superiority of our method.</div></div>","PeriodicalId":100790,"journal":{"name":"Journal of Information and Intelligence","volume":"3 6","pages":"Pages 481-491"},"PeriodicalIF":0.0,"publicationDate":"2025-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145665419","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-01DOI: 10.1016/j.jiixd.2025.07.003
Yong Zhang , Jingrui Zhang , Yanjie Dong , Feng Liang , Aohan Li , Xiping Hu
Federated learning (FL) has gained significant attention for enabling privacy preservation and knowledge sharing by transmitting model parameters from clients to a central server. However, with increasing network scale and limited bandwidth, uploading complete model parameters has become increasingly impractical. To address this challenge, we leverage the high informativeness of prototypes—feature centroids representing samples of the same class—and propose federated prototype momentum contrastive learning (FedPMC). At the communication level, FedPMC reduces communication overhead by using prototypes as carriers instead of full model parameters. At the local model update level, to mitigate overfitting, we construct an expanded batch sample space to incorporate richer visual information, design a supervised contrastive loss between global and real-time local prototypes, and adopt momentum contrast to gradually update the model. At the framework level, to fully exploit the sample's feature space, we employ three different pre-trained models for feature extraction and concatenate their outputs as input to the local model. FedPMC supports personalized local models and utilizes both global and local prototypes to address data heterogeneity among clients. We evaluate FedPMC alongside other state-of-the-art FL algorithms on the Digit-5 dataset within a unified lightweight framework to assess their comparative performance. The code is available at https://github.com/zhy665/fedPMC.
{"title":"Prototype-wise momentum-based federated contrast learning","authors":"Yong Zhang , Jingrui Zhang , Yanjie Dong , Feng Liang , Aohan Li , Xiping Hu","doi":"10.1016/j.jiixd.2025.07.003","DOIUrl":"10.1016/j.jiixd.2025.07.003","url":null,"abstract":"<div><div>Federated learning (FL) has gained significant attention for enabling privacy preservation and knowledge sharing by transmitting model parameters from clients to a central server. However, with increasing network scale and limited bandwidth, uploading complete model parameters has become increasingly impractical. To address this challenge, we leverage the high informativeness of prototypes—feature centroids representing samples of the same class—and propose federated prototype momentum contrastive learning (FedPMC). At the communication level, FedPMC reduces communication overhead by using prototypes as carriers instead of full model parameters. At the local model update level, to mitigate overfitting, we construct an expanded batch sample space to incorporate richer visual information, design a supervised contrastive loss between global and real-time local prototypes, and adopt momentum contrast to gradually update the model. At the framework level, to fully exploit the sample's feature space, we employ three different pre-trained models for feature extraction and concatenate their outputs as input to the local model. FedPMC supports personalized local models and utilizes both global and local prototypes to address data heterogeneity among clients. We evaluate FedPMC alongside other state-of-the-art FL algorithms on the Digit-5 dataset within a unified lightweight framework to assess their comparative performance. The code is available at <span><span>https://github.com/zhy665/fedPMC</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":100790,"journal":{"name":"Journal of Information and Intelligence","volume":"3 6","pages":"Pages 515-525"},"PeriodicalIF":0.0,"publicationDate":"2025-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145665462","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-11-01DOI: 10.1016/j.jiixd.2025.04.005
Peipei Zhao , Siyan Yang , Wei Ding , Ruyi Liu , Wentian Xin , Xiangzeng Liu , Qiguang Miao
Fine-grained visual classification (FGVC) is a very challenging task due to distinguishing subcategories under the same super-category. Recent works mainly localize discriminative image regions and capture subtle inter-class differences by utilizing attention-based methods. However, at the same layer, most attention-based works only consider large-scale attention blocks with the same size as feature maps, and they ignore small-scale attention blocks that are smaller than feature maps. To distinguish subcategories, it is important to exploit small local regions. In this work, a novel multi-scale attention network (MSANet) is proposed to capture large and small regions at the same layer in fine-grained visual classification. Specifically, a novel multi-scale attention layer (MSAL) is proposed, which generates multiple groups in each feature maps to capture different-scale discriminative regions. The groups based on large-scale regions can exploit global features and the groups based on the small-scale regions can extract local subtle features. Then, a simple feature fusion strategy is utilized to fully integrate global features and local subtle features to mine information that are more conducive to FGVC. Comprehensive experiments in Caltech-UCSD Birds-200-2011 (CUB), FGVC-Aircraft (AIR) and Stanford Cars (Cars) datasets show that our method achieves the competitive performances, which demonstrate its effectiveness.
{"title":"Learning multi-scale attention network for fine-grained visual classification","authors":"Peipei Zhao , Siyan Yang , Wei Ding , Ruyi Liu , Wentian Xin , Xiangzeng Liu , Qiguang Miao","doi":"10.1016/j.jiixd.2025.04.005","DOIUrl":"10.1016/j.jiixd.2025.04.005","url":null,"abstract":"<div><div>Fine-grained visual classification (FGVC) is a very challenging task due to distinguishing subcategories under the same super-category. Recent works mainly localize discriminative image regions and capture subtle inter-class differences by utilizing attention-based methods. However, at the same layer, most attention-based works only consider large-scale attention blocks with the same size as feature maps, and they ignore small-scale attention blocks that are smaller than feature maps. To distinguish subcategories, it is important to exploit small local regions. In this work, a novel multi-scale attention network (MSANet) is proposed to capture large and small regions at the same layer in fine-grained visual classification. Specifically, a novel multi-scale attention layer (MSAL) is proposed, which generates multiple groups in each feature maps to capture different-scale discriminative regions. The groups based on large-scale regions can exploit global features and the groups based on the small-scale regions can extract local subtle features. Then, a simple feature fusion strategy is utilized to fully integrate global features and local subtle features to mine information that are more conducive to FGVC. Comprehensive experiments in Caltech-UCSD Birds-200-2011 (CUB), FGVC-Aircraft (AIR) and Stanford Cars (Cars) datasets show that our method achieves the competitive performances, which demonstrate its effectiveness.</div></div>","PeriodicalId":100790,"journal":{"name":"Journal of Information and Intelligence","volume":"3 6","pages":"Pages 492-503"},"PeriodicalIF":0.0,"publicationDate":"2025-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145665460","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-01DOI: 10.1016/j.jiixd.2025.04.004
Mohammadreza Nakhaei , Reza Askari Moghadam
Supervised learning has been demonstrated to be a stable approach for training deep neural networks. Upside-down reinforcement learning solves reinforcement learning problems by using supervised learning, but this method suffers from weak sample efficiency in comparison to state-of-art reinforcement learning algorithms, mostly due to poor exploration. In this paper, we propose modifications to address this issue. To encourage better exploration, entropy maximization, noisy layer, and artificial curiosity are used in training upside-down reinforcement learning agents. Furthermore, to model sequences in reinforcement learning, recurrent neural networks are used in behavior function. We particularly propose deep clockwork RNN for this purpose. To prevent overfitting and underfitting due to a large or small number of updates respectively, we propose proportional number of updates according to the amount of new collected data instead of a fixed number in each iteration. This algorithm outperformed the original upside-down reinforcement learning and the results for several standard environments are presented.
{"title":"Improving sample efficiency and exploration in upside-down reinforcement learning","authors":"Mohammadreza Nakhaei , Reza Askari Moghadam","doi":"10.1016/j.jiixd.2025.04.004","DOIUrl":"10.1016/j.jiixd.2025.04.004","url":null,"abstract":"<div><div>Supervised learning has been demonstrated to be a stable approach for training deep neural networks. Upside-down reinforcement learning solves reinforcement learning problems by using supervised learning, but this method suffers from weak sample efficiency in comparison to state-of-art reinforcement learning algorithms, mostly due to poor exploration. In this paper, we propose modifications to address this issue. To encourage better exploration, entropy maximization, noisy layer, and artificial curiosity are used in training upside-down reinforcement learning agents. Furthermore, to model sequences in reinforcement learning, recurrent neural networks are used in behavior function. We particularly propose deep clockwork RNN for this purpose. To prevent overfitting and underfitting due to a large or small number of updates respectively, we propose proportional number of updates according to the amount of new collected data instead of a fixed number in each iteration. This algorithm outperformed the original upside-down reinforcement learning and the results for several standard environments are presented.</div></div>","PeriodicalId":100790,"journal":{"name":"Journal of Information and Intelligence","volume":"3 5","pages":"Pages 419-433"},"PeriodicalIF":0.0,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145371411","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-01DOI: 10.1016/j.jiixd.2025.06.003
Jingli Guo , Wenhao Liao , Dingzhang Guo , Ying Liu , Qiao Sun
A dual-band wide angle coverage cavity-backed slot antenna (WACCBSA) for Wi-Fi applications is proposed in this paper. The antenna's main structure consists of a metal cylindrical body featuring a horizontally oriented long slot on its side and an internally conformal thin backed cavity. The long slot, operating in a special mode, is excited by a strip within the thin cavity, enabling wide-angle coverage in the lower frequency band. The mechanism of this special thin cavity is analyzed. To enhance beam characteristics at higher frequencies, two parasitic elements are introduced to adjust the current distribution around the slot. In addition, a short slot is connected in parallel, 5 mm away from the long slot, to increase the working bandwidth. The measured results show that the proposed dual-band WACCBSA achieves two 10 dB reflection coefficient bandwidths of 13.97% (2.13 GHz∼2.45 GHz) and 11.25% (5.20 GHz∼5.82 GHz). Moreover, the measured gains of the proposed antenna exceed −4.9 dBi within ±90° in the horizontal plane for both frequency bands.
{"title":"A dual-band wide angle coverage cavity-backed slot antenna","authors":"Jingli Guo , Wenhao Liao , Dingzhang Guo , Ying Liu , Qiao Sun","doi":"10.1016/j.jiixd.2025.06.003","DOIUrl":"10.1016/j.jiixd.2025.06.003","url":null,"abstract":"<div><div>A dual-band wide angle coverage cavity-backed slot antenna (WACCBSA) for Wi-Fi applications is proposed in this paper. The antenna's main structure consists of a metal cylindrical body featuring a horizontally oriented long slot on its side and an internally conformal thin backed cavity. The long slot, operating in a special mode, is excited by a strip within the thin cavity, enabling wide-angle coverage in the lower frequency band. The mechanism of this special thin cavity is analyzed. To enhance beam characteristics at higher frequencies, two parasitic elements are introduced to adjust the current distribution around the slot. In addition, a short slot is connected in parallel, 5 mm away from the long slot, to increase the working bandwidth. The measured results show that the proposed dual-band WACCBSA achieves two 10 dB reflection coefficient bandwidths of 13.97% (2.13 GHz∼2.45 GHz) and 11.25% (5.20 GHz∼5.82 GHz). Moreover, the measured gains of the proposed antenna exceed −4.9 dBi within ±90° in the horizontal plane for both frequency bands.</div></div>","PeriodicalId":100790,"journal":{"name":"Journal of Information and Intelligence","volume":"3 5","pages":"Pages 453-462"},"PeriodicalIF":0.0,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145371413","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-01DOI: 10.1016/j.jiixd.2025.06.002
Yashuang Mu , Tian Liu , Hongyue Guo , Xianchao Zhu , Lidong Wang , Benhang Liu , Linlin Guo
Time series segmentation aims to extract some meaningful subsequences from complex temporal information. A proper segmentation can effectively help users to analyze the structure of time series. In this study, we propose an information granulation-based fuzzy clustering method for the problem of time series segmentation. The suggested time series segmentation method follows the technological procedure of fuzzy c-means clustering method. First, the original time series is randomly divided into several segments. Then, an information granulation-based dynamic time warping approach is designed to update the series centers, where the principle of reasonable granularity is utilized to calculate the mean of the segments. Next, the time series segments are clustered by optimizing the objective function. Finally, the optimal segmentation points are generated by merging the contiguous segments in the same cluster. The experimental results show that the established segmentation method has more advantages than the existing segmentation methods.
{"title":"An information granulation-based fuzzy clustering method for time series segmentation","authors":"Yashuang Mu , Tian Liu , Hongyue Guo , Xianchao Zhu , Lidong Wang , Benhang Liu , Linlin Guo","doi":"10.1016/j.jiixd.2025.06.002","DOIUrl":"10.1016/j.jiixd.2025.06.002","url":null,"abstract":"<div><div>Time series segmentation aims to extract some meaningful subsequences from complex temporal information. A proper segmentation can effectively help users to analyze the structure of time series. In this study, we propose an information granulation-based fuzzy clustering method for the problem of time series segmentation. The suggested time series segmentation method follows the technological procedure of fuzzy c-means clustering method. First, the original time series is randomly divided into several segments. Then, an information granulation-based dynamic time warping approach is designed to update the series centers, where the principle of reasonable granularity is utilized to calculate the mean of the segments. Next, the time series segments are clustered by optimizing the objective function. Finally, the optimal segmentation points are generated by merging the contiguous segments in the same cluster. The experimental results show that the established segmentation method has more advantages than the existing segmentation methods.</div></div>","PeriodicalId":100790,"journal":{"name":"Journal of Information and Intelligence","volume":"3 5","pages":"Pages 434-452"},"PeriodicalIF":0.0,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145371412","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-01DOI: 10.1016/j.jiixd.2024.10.001
Kun Cheng , Mingrui Zhu , Nannan Wang , Xinbo Gao
Benefiting from the advancement of deep learning techniques, face photo-sketch synthesis has witnessed significant progress in recent years. Cutting-edge methods typically treat this task as an image-to-image translation problem and train a conditional generative model to learn the mapping between two domains. However, purely parametric deep learning models often struggle to capture instance-level details due to limited training samples and tend to focus on domain-level mapping. Moreover, sketch-to-photo synthesis is more challenging than photo-to-sketch synthesis and holds greater significance in the realm of public security, but it has not been well-studied in existing methods. To address these challenges, we introduce an innovative framework that synergistically integrates parametric and non-parametric approaches, infusing facial generative priors and instance-level prior knowledge from the target domain to enrich texture detail synthesis. Specifically, our framework employs a semantic-aware network to facilitate coarse cross-domain reconstruction, thereby capturing domain-level information. Moreover, through efficient neural patch matching between the input image and multiple reference (training) samples, we can harness instance-level prior knowledge as a detailed texture representation to enhance detail fidelity. For the sketch-to-photo synthesis task, we further propose a local patch correspondence mechanism that improves the rationality of matching through local constraint. To further enhance the generation of realistic and detailed facial features, we incorporate a pre-trained StyleGAN as the decoder, leveraging its extensive facial generative priors. Additionally, we introduce the relaxed Earth Movers Distance (rEMD) loss to improve the style consistency between the generated results and the target domain. Extensive experiments show that our method achieves state-of-the-art performance on both quantitative and qualitative evaluations.
{"title":"Integrating hybrid priors for face photo-sketch synthesis","authors":"Kun Cheng , Mingrui Zhu , Nannan Wang , Xinbo Gao","doi":"10.1016/j.jiixd.2024.10.001","DOIUrl":"10.1016/j.jiixd.2024.10.001","url":null,"abstract":"<div><div>Benefiting from the advancement of deep learning techniques, face photo-sketch synthesis has witnessed significant progress in recent years. Cutting-edge methods typically treat this task as an image-to-image translation problem and train a conditional generative model to learn the mapping between two domains. However, purely parametric deep learning models often struggle to capture instance-level details due to limited training samples and tend to focus on domain-level mapping. Moreover, sketch-to-photo synthesis is more challenging than photo-to-sketch synthesis and holds greater significance in the realm of public security, but it has not been well-studied in existing methods. To address these challenges, we introduce an innovative framework that synergistically integrates parametric and non-parametric approaches, infusing facial generative priors and instance-level prior knowledge from the target domain to enrich texture detail synthesis. Specifically, our framework employs a semantic-aware network to facilitate coarse cross-domain reconstruction, thereby capturing domain-level information. Moreover, through efficient neural patch matching between the input image and multiple reference (training) samples, we can harness instance-level prior knowledge as a detailed texture representation to enhance detail fidelity. For the sketch-to-photo synthesis task, we further propose a local patch correspondence mechanism that improves the rationality of matching through local constraint. To further enhance the generation of realistic and detailed facial features, we incorporate a pre-trained StyleGAN as the decoder, leveraging its extensive facial generative priors. Additionally, we introduce the relaxed Earth Movers Distance (rEMD) loss to improve the style consistency between the generated results and the target domain. Extensive experiments show that our method achieves state-of-the-art performance on both quantitative and qualitative evaluations.</div></div>","PeriodicalId":100790,"journal":{"name":"Journal of Information and Intelligence","volume":"3 5","pages":"Pages 401-418"},"PeriodicalIF":0.0,"publicationDate":"2025-09-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145371410","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","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