Multimodal knowledge graphs (MMKGs) have gained widespread adoption across various domains. However, existing transformer-based methods for MMKG representation learning primarily focus on enhancing representation performance, while overlooking time and memory costs, which reduces model efficiency. To tackle these limitations, we introduce a multimodal lightweight transformer (MLFormer) model, which not only ensures robust representation capabilities but also considerably improves computational efficiency. We find that the self-attention mechanism in transformers leads to substantial performance overheads. As a result, we optimize the traditional MMKGE model in two aspects: modality processing and modality fusion, by incorporating a filter gate and Fourier transform. Our experimental results on real-world multimodal knowledge graph completion datasets demonstrate that MLFormer achieves significant improvements in computational efficiency while maintaining competitive performance.
{"title":"MLFormer: Unleashing Efficiency Without Attention for Multimodal Knowledge Graph Embedding","authors":"Meng Wang;Changyu Li;Feiyu Chen;Jie Shao;Ke Qin;Shuang Liang","doi":"10.1109/TCSS.2025.3620089","DOIUrl":"https://doi.org/10.1109/TCSS.2025.3620089","url":null,"abstract":"Multimodal knowledge graphs (MMKGs) have gained widespread adoption across various domains. However, existing transformer-based methods for MMKG representation learning primarily focus on enhancing representation performance, while overlooking time and memory costs, which reduces model efficiency. To tackle these limitations, we introduce a multimodal lightweight transformer (MLFormer) model, which not only ensures robust representation capabilities but also considerably improves computational efficiency. We find that the self-attention mechanism in transformers leads to substantial performance overheads. As a result, we optimize the traditional MMKGE model in two aspects: modality processing and modality fusion, by incorporating a filter gate and Fourier transform. Our experimental results on real-world multimodal knowledge graph completion datasets demonstrate that MLFormer achieves significant improvements in computational efficiency while maintaining competitive performance.","PeriodicalId":13044,"journal":{"name":"IEEE Transactions on Computational Social Systems","volume":"12 6","pages":"5536-5549"},"PeriodicalIF":4.5,"publicationDate":"2025-11-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145652138","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-27DOI: 10.1109/TCSS.2025.3614707
Jingjing Shi;Zhihua Guo;Yumei Huang
As the behavior of user comment brushing on e-commerce and social platforms becomes increasingly hidden, this article constructs a detection algorithm that integrates dynamic graph neural network (dynamic GNN) and federated learning to detect the blind spots caused by deep learning-generated text and cross-platform collaborative brushing. Dynamic GNN is used to model user-device temporal associations to identify group topological features, and federated learning is used to aggregate multiplatform features to improve cross-platform detection performance while protecting privacy. Based on user comment behavior sequences, such as device ID (identifier), IP (Internet protocol), and timestamp, a dynamic heterogeneous graph (nodes: users/devices; edges: interaction frequency and time series) is constructed, and the topological structure is updated through a sliding window to capture short-term collaborative brushing patterns. A time-aware graph attention mechanism is adopted to aggregate the historical states of neighbor nodes and the current interaction features and output the temporal embedding vector of the user node to characterize its membership in the brushing group. Each platform trains the dynamic GNN model locally, and the central server aggregates cross-platform features such as device fingerprints and IP geographic distribution through federated averaging (FedAvg) to avoid the sharing of raw data. The user temporal embedding is concatenated with the federated features and input into the multilayer perceptron (MLP). The probability of user brushing is output, and the suspicious groups are marked after the threshold is determined. Experimental results show that the dynamic GNN integrated with federated learning has a false alarm rate of 12.1% and an F1-score of 83.1% under an attack density of 50%, demonstrating high cross-platform detection performance. When the time window changes from 30 to 600 s, the mean feature update delay decreases linearly with the increase of the window (38.2→15.9 ms), maintaining a millisecond-level response. The changing trend of the mean training throughput (12 450→29 450 edges/s) directly reflects the elastic expansion capability of the model architecture and has a high dynamic topology capture timeliness. The experimental data verify the effectiveness of this article’s research on the algorithm for identifying user comment brushing behavior.
{"title":"User Comment Brushing Behavior Identification Algorithm for Malicious Network Behavior Detection","authors":"Jingjing Shi;Zhihua Guo;Yumei Huang","doi":"10.1109/TCSS.2025.3614707","DOIUrl":"https://doi.org/10.1109/TCSS.2025.3614707","url":null,"abstract":"As the behavior of user comment brushing on e-commerce and social platforms becomes increasingly hidden, this article constructs a detection algorithm that integrates dynamic graph neural network (dynamic GNN) and federated learning to detect the blind spots caused by deep learning-generated text and cross-platform collaborative brushing. Dynamic GNN is used to model user-device temporal associations to identify group topological features, and federated learning is used to aggregate multiplatform features to improve cross-platform detection performance while protecting privacy. Based on user comment behavior sequences, such as device ID (identifier), IP (Internet protocol), and timestamp, a dynamic heterogeneous graph (nodes: users/devices; edges: interaction frequency and time series) is constructed, and the topological structure is updated through a sliding window to capture short-term collaborative brushing patterns. A time-aware graph attention mechanism is adopted to aggregate the historical states of neighbor nodes and the current interaction features and output the temporal embedding vector of the user node to characterize its membership in the brushing group. Each platform trains the dynamic GNN model locally, and the central server aggregates cross-platform features such as device fingerprints and IP geographic distribution through federated averaging (FedAvg) to avoid the sharing of raw data. The user temporal embedding is concatenated with the federated features and input into the multilayer perceptron (MLP). The probability of user brushing is output, and the suspicious groups are marked after the threshold is determined. Experimental results show that the dynamic GNN integrated with federated learning has a false alarm rate of 12.1% and an F1-score of 83.1% under an attack density of 50%, demonstrating high cross-platform detection performance. When the time window changes from 30 to 600 s, the mean feature update delay decreases linearly with the increase of the window (38.2→15.9 ms), maintaining a millisecond-level response. The changing trend of the mean training throughput (12 450→29 450 edges/s) directly reflects the elastic expansion capability of the model architecture and has a high dynamic topology capture timeliness. The experimental data verify the effectiveness of this article’s research on the algorithm for identifying user comment brushing behavior.","PeriodicalId":13044,"journal":{"name":"IEEE Transactions on Computational Social Systems","volume":"13 1","pages":"1180-1193"},"PeriodicalIF":4.5,"publicationDate":"2025-10-27","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146175839","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-17DOI: 10.1109/TCSS.2025.3609251
Fan Gao;Himanshu Dhumras;Garima Thakur;Xingsi Xue;Ya-Juan Yang
The exponential proliferation of social media and Internet of Things (IoT) technologies has paved the way for transformative applications in public health, particularly for the early detection of mental health concerns. This study introduces an innovative framework leveraging cooperative learning algorithms combined with intelligence of social things (IoST) data to enhance mental health issue detection. By integrating multimodal user data from social platforms, wearable devices, and IoT sensors, the proposed approach achieves superior predictive accuracy, with the random forest-based model outperforming benchmarks at 88% accuracy and a 0.90 receiver operating characteristic area under the curve (ROC-AUC). The incorporation of key features, including social homophily and real-time behavioral metrics, significantly bolsters detection rates. Ethical considerations, including data privacy and bias reduction, are meticulously addressed, ensuring a scalable and user-centered solution. The findings underscore the potential of IoST-driven cooperative algorithms to revolutionize mental health interventions by enabling timely, precise, and ethical detection systems.
{"title":"Leveraging Cooperative Learning Algorithms for Early Detection of Mental Health Issues Using Intelligence of Social Things Data","authors":"Fan Gao;Himanshu Dhumras;Garima Thakur;Xingsi Xue;Ya-Juan Yang","doi":"10.1109/TCSS.2025.3609251","DOIUrl":"https://doi.org/10.1109/TCSS.2025.3609251","url":null,"abstract":"The exponential proliferation of social media and Internet of Things (IoT) technologies has paved the way for transformative applications in public health, particularly for the early detection of mental health concerns. This study introduces an innovative framework leveraging cooperative learning algorithms combined with intelligence of social things (IoST) data to enhance mental health issue detection. By integrating multimodal user data from social platforms, wearable devices, and IoT sensors, the proposed approach achieves superior predictive accuracy, with the random forest-based model outperforming benchmarks at 88% accuracy and a 0.90 receiver operating characteristic area under the curve (ROC-AUC). The incorporation of key features, including social homophily and real-time behavioral metrics, significantly bolsters detection rates. Ethical considerations, including data privacy and bias reduction, are meticulously addressed, ensuring a scalable and user-centered solution. The findings underscore the potential of IoST-driven cooperative algorithms to revolutionize mental health interventions by enabling timely, precise, and ethical detection systems.","PeriodicalId":13044,"journal":{"name":"IEEE Transactions on Computational Social Systems","volume":"13 1","pages":"1091-1099"},"PeriodicalIF":4.5,"publicationDate":"2025-10-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146175683","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Functional near-infrared spectroscopy (fNIRS) is widely used to monitor cognitive load, yet classification in dynamic, cooperative settings remains challenging due to noise and hemodynamic delays. This article aims to develop a biologically inspired classification framework for fNIRS data that is suitable for both individual and collaborative learning environments. We propose fNIRS-SpikeNet, a spiking neural network (SNN) framework that integrates rate, latency, and delta spike encoding strategies with a residual-SNN architecture to capture spatiotemporal dynamics. We evaluate our method on three public fNIRS datasets involving mental and motor tasks. Experimental results demonstrate that fNIRS-SpikeNet, particularly under rate encoding, significantly outperforms conventional machine learning and deep learning baselines in accuracy, efficiency, and real-time adaptability. These outcomes highlight the potential of SNNs for low-power, real-time neuroimaging in socially interactive applications.
{"title":"fNIRS-SpikeNet: A Spiking Neural Network Framework for Cognitive Load Classification in Cooperative Learning Environments","authors":"Peijiang Zhang;Tao Cheng;Yuande Jiang;Xiaochuan Zou;Xiaoming Chen","doi":"10.1109/TCSS.2025.3598044","DOIUrl":"https://doi.org/10.1109/TCSS.2025.3598044","url":null,"abstract":"Functional near-infrared spectroscopy (fNIRS) is widely used to monitor cognitive load, yet classification in dynamic, cooperative settings remains challenging due to noise and hemodynamic delays. This article aims to develop a biologically inspired classification framework for fNIRS data that is suitable for both individual and collaborative learning environments. We propose fNIRS-SpikeNet, a spiking neural network (SNN) framework that integrates rate, latency, and delta spike encoding strategies with a residual-SNN architecture to capture spatiotemporal dynamics. We evaluate our method on three public fNIRS datasets involving mental and motor tasks. Experimental results demonstrate that fNIRS-SpikeNet, particularly under rate encoding, significantly outperforms conventional machine learning and deep learning baselines in accuracy, efficiency, and real-time adaptability. These outcomes highlight the potential of SNNs for low-power, real-time neuroimaging in socially interactive applications.","PeriodicalId":13044,"journal":{"name":"IEEE Transactions on Computational Social Systems","volume":"13 1","pages":"1134-1145"},"PeriodicalIF":4.5,"publicationDate":"2025-10-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146175674","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-10-06DOI: 10.1109/TCSS.2025.3608423
{"title":"IEEE Transactions on Computational Social Systems Information for Authors","authors":"","doi":"10.1109/TCSS.2025.3608423","DOIUrl":"https://doi.org/10.1109/TCSS.2025.3608423","url":null,"abstract":"","PeriodicalId":13044,"journal":{"name":"IEEE Transactions on Computational Social Systems","volume":"12 5","pages":"C4-C4"},"PeriodicalIF":4.5,"publicationDate":"2025-10-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11194050","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145230025","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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 : 2025-10-06DOI: 10.1109/TCSS.2025.3608419
{"title":"IEEE Transactions on Computational Social Systems Publication Information","authors":"","doi":"10.1109/TCSS.2025.3608419","DOIUrl":"https://doi.org/10.1109/TCSS.2025.3608419","url":null,"abstract":"","PeriodicalId":13044,"journal":{"name":"IEEE Transactions on Computational Social Systems","volume":"12 5","pages":"C2-C2"},"PeriodicalIF":4.5,"publicationDate":"2025-10-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11194049","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145315350","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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 : 2025-10-06DOI: 10.1109/TCSS.2025.3606570
Amit Kumar Singh;Jungong Han;Stefano Berretti
{"title":"Guest Editorial: Special Issue on Trends in Social Multimedia Computing: Models, Methodologies, and Applications","authors":"Amit Kumar Singh;Jungong Han;Stefano Berretti","doi":"10.1109/TCSS.2025.3606570","DOIUrl":"https://doi.org/10.1109/TCSS.2025.3606570","url":null,"abstract":"","PeriodicalId":13044,"journal":{"name":"IEEE Transactions on Computational Social Systems","volume":"12 5","pages":"3747-3750"},"PeriodicalIF":4.5,"publicationDate":"2025-10-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11193968","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145230028","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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 : 2025-10-06DOI: 10.1109/TCSS.2025.3608421
{"title":"IEEE Systems, Man, and Cybernetics Society Information","authors":"","doi":"10.1109/TCSS.2025.3608421","DOIUrl":"https://doi.org/10.1109/TCSS.2025.3608421","url":null,"abstract":"","PeriodicalId":13044,"journal":{"name":"IEEE Transactions on Computational Social Systems","volume":"12 5","pages":"C3-C3"},"PeriodicalIF":4.5,"publicationDate":"2025-10-06","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=11193967","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145230019","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"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 : 2025-09-22DOI: 10.1109/TCSS.2025.3600391
Zhaodi Wang;Yangyan Zeng;Biao Leng;Xiaokang Zhou
Scene graph generation (SGG) plays an important role in the intelligence of social things (IoST) framework by extracting structured semantic representations from social device data, thereby supporting advanced scene understanding and behavioral-cultural modeling. However, the intrinsic long-tail nature of real-world social device data, coupled with the semantic entanglement between head and tail categories (e.g., “on” versus “standing on”), presents significant challenges for fine-grained SGG. This often results in biased models and suboptimal generalization to rare but semantically informative relations. To address these issues, we propose a novel cooperative dual classifier (CDC) framework for fine-grained SGG in IoST-driven social systems. CDC introduces a cooperative learning mechanism that combines two classifiers. The frozen prototype classifier is designed with maximum interclass margins to alleviate class imbalance. In parallel, a learnable classifier dynamically adjusts decision boundaries to improve discriminative precision. To further enhance the integration between the two classifiers, we introduce a weight knowledge transfer (WKT) module and a collaborative constraint term, facilitating robust adaptation to tail categories. Extensive experiments on the Visual Genome and GQA datasets demonstrate that CDC outperforms state-of-the-art SGG methods, particularly in modeling fine-grained relations under long-tail distributions. These results highlight the capability of CDC to advance semantic understanding of complex behavioral and cultural patterns within computational social systems.
{"title":"CDC: Enhancing Scene Graph Generation for IoST-Driven Social Behavioral Modeling With Cooperative Dual Classifier","authors":"Zhaodi Wang;Yangyan Zeng;Biao Leng;Xiaokang Zhou","doi":"10.1109/TCSS.2025.3600391","DOIUrl":"https://doi.org/10.1109/TCSS.2025.3600391","url":null,"abstract":"Scene graph generation (SGG) plays an important role in the intelligence of social things (IoST) framework by extracting structured semantic representations from social device data, thereby supporting advanced scene understanding and behavioral-cultural modeling. However, the intrinsic long-tail nature of real-world social device data, coupled with the semantic entanglement between head and tail categories (e.g., “on” versus “standing on”), presents significant challenges for fine-grained SGG. This often results in biased models and suboptimal generalization to rare but semantically informative relations. To address these issues, we propose a novel cooperative dual classifier (CDC) framework for fine-grained SGG in IoST-driven social systems. CDC introduces a cooperative learning mechanism that combines two classifiers. The frozen prototype classifier is designed with maximum interclass margins to alleviate class imbalance. In parallel, a learnable classifier dynamically adjusts decision boundaries to improve discriminative precision. To further enhance the integration between the two classifiers, we introduce a weight knowledge transfer (WKT) module and a collaborative constraint term, facilitating robust adaptation to tail categories. Extensive experiments on the Visual Genome and GQA datasets demonstrate that CDC outperforms state-of-the-art SGG methods, particularly in modeling fine-grained relations under long-tail distributions. These results highlight the capability of CDC to advance semantic understanding of complex behavioral and cultural patterns within computational social systems.","PeriodicalId":13044,"journal":{"name":"IEEE Transactions on Computational Social Systems","volume":"13 1","pages":"1120-1133"},"PeriodicalIF":4.5,"publicationDate":"2025-09-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146175930","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","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：