Information Fusion最新文献_第2页

Unsupervised feature selection via unifying distribution alignment and structure preservation 统一分布对齐和结构保存的无监督特征选择

IF 15.5 1区计算机科学

Information Fusion Pub Date : 2025-07-26 DOI: 10.1016/j.inffus.2025.103544

Chunxu Cao , Yong Zhang , Yanke Ai , Qiang Zhang

{"title":"Unsupervised feature selection via unifying distribution alignment and structure preservation","authors":"Chunxu Cao , Yong Zhang , Yanke Ai , Qiang Zhang","doi":"10.1016/j.inffus.2025.103544","DOIUrl":"10.1016/j.inffus.2025.103544","url":null,"abstract":"<div><div>The increasing complexity of high-dimensional data demands effective feature selection techniques that preserve both global distributional characteristics and local structures. However, existing methods often encounter a fundamental trade-off between preserving global distributional fidelity and maintaining local geometric structures, leading to information loss. This work presents a novel kernel-enhanced Gromov–Wasserstein alignment framework that unifies global distribution alignment and local structure preservation. Our approach leverages Gromov–Wasserstein distance and the kernel trick to enhance metric space comparisons, effectively capturing nonlinear relationships while improving stability in noisy data. To ensure scalability, we develop an efficient randomized filter algorithm, balancing computational efficiency with feature diversity. Extensive experiments across 20 benchmark datasets demonstrate the superior performance of our method, showing that it surpasses state-of-the-art feature selection techniques. These results highlight the effectiveness of integrating distributional alignment and structure preservation for unsupervised feature selection in high-dimensional data analysis.</div></div>","PeriodicalId":50367,"journal":{"name":"Information Fusion","volume":"126 ","pages":"Article 103544"},"PeriodicalIF":15.5,"publicationDate":"2025-07-26","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144724760","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Urban data fusion for spatio-temporal incident forecasting using graph attention and generative AI 基于图注意力和生成式人工智能的城市数据融合时空事件预测

IF 15.5 1区计算机科学

Information Fusion Pub Date : 2025-07-25 DOI: 10.1016/j.inffus.2025.103532

Niranchana Radhakrishnan , Hemalatha S. , Ganesh Gopal Devarajan , Nachiyappan S. , Karthick S. , Abhishek Singhal

{"title":"Urban data fusion for spatio-temporal incident forecasting using graph attention and generative AI","authors":"Niranchana Radhakrishnan , Hemalatha S. , Ganesh Gopal Devarajan , Nachiyappan S. , Karthick S. , Abhishek Singhal","doi":"10.1016/j.inffus.2025.103532","DOIUrl":"10.1016/j.inffus.2025.103532","url":null,"abstract":"<div><div>Accuracy and interpretability are two essential properties for any predictive model built on complex multi-source spatio-temporal data. In domains such as urban dynamics, environmental monitoring, and intelligent transportation systems, predictive models must effectively integrate heterogeneous data sources while maintaining transparency in decision making. This paper introduces a novel Two-Stage Attention Integrated Graph-Based Multi-source Spatio-Temporal Data Fusion Network (2S-AGMSTDF) that leverages generative AI-enhanced representations to improve both prediction accuracy and model interpretability. In the first stage, we perform external feature embedding using an Augmented Knowledge Graph Convolutional Network (AKGCN), spatial feature embedding via a GCN-ResNet-based Cross-Modality Transformer (GRCMT), and temporal feature embedding using an Attention-Based LSTM. The second stage fuses the learned embeddings through a Sparse Attentive Backtracking LSTM, enabling the model to trace influential patterns across modalities and time steps. This architecture allows for effective fusion of diverse data types, such as environmental indicators, mobility patterns, and auxiliary data such as traffic flow and point-of-interest (POI) information. The proposed method demonstrates how generative AI techniques can improve both the quality of the representation and the fusion process across modalities, enabling scalable and interpretable predictions in complex real-world scenarios. Extensive experiments on real-world urban datasets validate the superiority of 2S-AGMSTDF in terms of predictive performance and model transparency, making it a strong candidate for GenAI-based information fusion applications.</div></div>","PeriodicalId":50367,"journal":{"name":"Information Fusion","volume":"126 ","pages":"Article 103532"},"PeriodicalIF":15.5,"publicationDate":"2025-07-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144738912","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

XpertDx: Expert-level diabetic retinopathy lesion segmentation with cross-domain feature fusion XpertDx：基于跨域特征融合的专家级糖尿病视网膜病变分割

IF 14.7 1区计算机科学

Information Fusion Pub Date : 2025-07-25 DOI: 10.1016/j.inffus.2025.103556

Fei Ma , Guangmei Jia , Fen Yan , Yuefeng Ma , Ronghua Cheng , Jing Meng

{"title":"XpertDx: Expert-level diabetic retinopathy lesion segmentation with cross-domain feature fusion","authors":"Fei Ma , Guangmei Jia , Fen Yan , Yuefeng Ma , Ronghua Cheng , Jing Meng","doi":"10.1016/j.inffus.2025.103556","DOIUrl":"10.1016/j.inffus.2025.103556","url":null,"abstract":"<div><div>Diabetic retinopathy is a microvascular disease that poses a significant threat to visual health, and its automatic segmentation is crucial for early diagnosis and intervention. Optical Coherence Tomography Angiography (OCTA) is a non-invasive imaging technique capable of obtaining high-resolution structures of retinal and choroidal vasculature. However, due to the minimal blood flow changes associated with early microlesions, these subtle abnormalities are often overlooked in imaging. Furthermore, traditional segmentation methods primarily rely on information from a single perspective provided by a single network, making it challenging to effectively capture the complex characteristics of such lesions. To address these issues, we propose a novel lesion segmentation framework for the precise segmentation of retinal lesion regions in OCTA images. Specifically, we design a frequency-domain encoder with multi-level discrete wavelet transform to capture multi-scale texture features. An adaptive fusion perception module (AFPM) is then employed to facilitate deep interaction and alignment between spatial and frequency domain features. In addition, we developed a comparative monitoring module that embeds a contrastive learning mechanism at the patch level. Furthermore, we propose a consistency learning strategy with multi-path decoding and consistency correction to capture details in complex lesion regions. Experimental results on two OCTA DR datasets show that our method outperforms existing state-of-the-art methods. This consolidates that the analysis of retinal lesions may offer a new scheme for the study of various neurodegenerative diseases.</div></div>","PeriodicalId":50367,"journal":{"name":"Information Fusion","volume":"126 ","pages":"Article 103556"},"PeriodicalIF":14.7,"publicationDate":"2025-07-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144712007","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Consensus-based distributed secure state estimation for cyber-physical systems against random cyber attacks 针对随机网络攻击的基于共识的网络物理系统分布式安全状态估计

IF 15.5 1区计算机科学

Information Fusion Pub Date : 2025-07-25 DOI: 10.1016/j.inffus.2025.103545

Zhichen Han , Shengbing Zhang , Zengwang Jin , Changyin Sun

{"title":"Consensus-based distributed secure state estimation for cyber-physical systems against random cyber attacks","authors":"Zhichen Han , Shengbing Zhang , Zengwang Jin , Changyin Sun","doi":"10.1016/j.inffus.2025.103545","DOIUrl":"10.1016/j.inffus.2025.103545","url":null,"abstract":"<div><div>Security issues of cyber–physical systems (CPSs) have attracted extensive attention due to the increasing vulnerability to malicious threats from cyber attacks. This paper focuses on the problem of consensus-based distributed secure state estimation for CPSs against cyber attacks. A novel cyber attack mode is established with high randomness. Specifically, the attack is randomly set up in attack position, type, duration and value, and can cause significant devastation due to the stochastic and unpredictable attack characteristics. To counter such attack behavior, a consensus-based distributed secure state estimation with a compensation-based measurement is proposed. The compensation-based measurement, obtained by combining an attack detector and prediction measurement, is utilized to replace the current measurement in the attacked channel. In the consensus-based distributed secure state estimation algorithm, the distributed estimators first adopt the compensation-based measurement to perform the local information fusion estimation, and then use the transmitted estimates to realize the consensus estimation. Furthermore, the information fusion estimator is recursively derived, and a sufficient condition is provided based on the Lyapunov method to provide the convergence of the consensus estimation. Finally, a series of numerical simulations are provided to illustrate the effectiveness of the proposed method.</div></div>","PeriodicalId":50367,"journal":{"name":"Information Fusion","volume":"126 ","pages":"Article 103545"},"PeriodicalIF":15.5,"publicationDate":"2025-07-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144724816","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

MVG-Splatting: Multi-view guided Gaussian Splatting with adaptive quantile-based geometric consistency densification MVG-Splatting：基于自适应分位数的几何一致性密度的多视图引导高斯飞溅

IF 14.7 1区计算机科学

Information Fusion Pub Date : 2025-07-24 DOI: 10.1016/j.inffus.2025.103540

Zhuoxiao Li , Shanliang Yao , Yijie Chu , Ángel F. García-Fernández , Yong Yue , Weiping Ding , Xiaohui Zhu

{"title":"MVG-Splatting: Multi-view guided Gaussian Splatting with adaptive quantile-based geometric consistency densification","authors":"Zhuoxiao Li , Shanliang Yao , Yijie Chu , Ángel F. García-Fernández , Yong Yue , Weiping Ding , Xiaohui Zhu","doi":"10.1016/j.inffus.2025.103540","DOIUrl":"10.1016/j.inffus.2025.103540","url":null,"abstract":"<div><div>In the rapidly evolving field of image-fusion-based 3D reconstruction, 3D Gaussian Splatting (3DGS) and 2D Gaussian Splatting (2DGS) represent significant advancements. Although 2DGS compresses 3D Gaussian primitives into 2D Gaussian surfels to effectively enhance mesh extraction quality, this compression can potentially lead to a decrease in rendering quality. Additionally, unreliable densification processes and the calculation of depth through the accumulation of opacity can compromise the detail of mesh extraction. Specifically, we integrate an optimized method for calculating normals, which, combined with image gradients, helps rectify inconsistencies in the original depth computations. Additionally, utilizing projection strategies akin to those in Multi-View Stereo (MVS), we propose an adaptive quantile-based method that dynamically determines the level of additional densification guided by depth maps, from coarse to fine detail. Furthermore, we design a joint loss function that combines edge-aware and feature-aware depth constraints to ensure that our refined depth aligns well with the ground-truth image edges and features, thereby improving both photometric and geometric consistency. Experimental evidence demonstrates that our method not only resolves the issues of rendering quality degradation caused by depth discrepancies but also facilitates direct mesh extraction from denser Gaussian point clouds using the Marching Cubes algorithm. This approach significantly enhances the overall fidelity and accuracy of the 3D reconstruction process, ensuring that both the geometric details and visual quality. The project is available at <span><span>https://mvgsplatting.github.io/</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50367,"journal":{"name":"Information Fusion","volume":"126 ","pages":"Article 103540"},"PeriodicalIF":14.7,"publicationDate":"2025-07-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144702801","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

ST-Camba: A decoupled-free spatiotemporal graph fusion state space model with linear complexity for efficient traffic forecasting ST-Camba：一种具有线性复杂度的无解耦时空图融合状态空间模型，用于有效的交通预测

IF 14.7 1区计算机科学

Information Fusion Pub Date : 2025-07-24 DOI: 10.1016/j.inffus.2025.103495

Xiangxu Wang , Jinzhou Cao , Tianhong Zhao , Bowen Zhang , Guanzhou Chen , Zhenhui Li , Haolin Chen , Wei Tu , Qingquan Li

{"title":"ST-Camba: A decoupled-free spatiotemporal graph fusion state space model with linear complexity for efficient traffic forecasting","authors":"Xiangxu Wang , Jinzhou Cao , Tianhong Zhao , Bowen Zhang , Guanzhou Chen , Zhenhui Li , Haolin Chen , Wei Tu , Qingquan Li","doi":"10.1016/j.inffus.2025.103495","DOIUrl":"10.1016/j.inffus.2025.103495","url":null,"abstract":"<div><div>Traffic forecasting is a critical task in intelligent transportation systems, requiring accurate modeling of spatiotemporal dependencies among traffic sensors. Traditional deep-learning methods face two key challenges: (1) decoupled spatial–temporal pipelines that process and fuse spatial and temporal dimensions separately fail to capture their intricate interdependencies; and (2) state-of-the-art (SOTA) models relying on Transformer architectures often struggle to balance computational efficiency with representational capacity. To address these limitations, we propose ST-Camba, a novel decoupled-free spatiotemporal graph fusion state space model that unifies spatial and temporal dimensions within a single framework. ST-Camba is the first to integrate a spatial dimension axis into state space equations, enabling effective coupled spatiotemporal modeling through graph convolutions while inheriting the linear complexity advantage of Mamba series models. Additionally, we design an Adaptive Spatial Structure (ASS) Injector and a Lerp-based Gated Unit (LGU) to facilitate adaptive spatial structure capture and control information flow in spatiotemporal modeling. Extensive experiments on flow and speed prediction tasks across standard datasets demonstrate ST-Camba’s superiority. Specifically, on the PEMS07 dataset, our model achieves a 1.8% reduction in MAE compared to other baselines, while reducing computational costs by up to 14.5%. This work underscores the necessity of coupled spatiotemporal modeling and provides a theoretical foundation for scalable solutions in urban traffic systems.</div></div>","PeriodicalId":50367,"journal":{"name":"Information Fusion","volume":"126 ","pages":"Article 103495"},"PeriodicalIF":14.7,"publicationDate":"2025-07-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144712951","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

DeepTurbid:Underwater marker detection and pose estimation in turbid conditions DeepTurbid：在浑浊条件下水下标记检测和姿态估计

IF 15.5 1区计算机科学

Information Fusion Pub Date : 2025-07-23 DOI: 10.1016/j.inffus.2025.103568

Fanyi Meng , Zheng Cong , Bing Wang , Kai Guo , Qingquan Li , Dejin Zhang

{"title":"DeepTurbid:Underwater marker detection and pose estimation in turbid conditions","authors":"Fanyi Meng , Zheng Cong , Bing Wang , Kai Guo , Qingquan Li , Dejin Zhang","doi":"10.1016/j.inffus.2025.103568","DOIUrl":"10.1016/j.inffus.2025.103568","url":null,"abstract":"<div><div>Using fiducial markers for visual localization provides a consistent solution for underwater positioning. However, image degradation caused by water turbidity often leads to the failure of traditional, hand-crafted detection methods. To address this challenge, we introduce DeepTurbid, a system for underwater marker detection and pose estimation. Building on existing ArUco marker systems, DeepTurbid defines the mapping of marker control points in challenging underwater environments and employs a high-resolution neural network for keypoint prediction and marker ID decoding. Moreover, by leveraging the underwater imaging model and the optical properties of underwater scenes, we propose an underwater marker image generation scheme and an adaptive heatmap labeling. This approach generates a diverse marker image dataset, spanning multiple water types and degradation levels for network training. We evaluate DeepTurbid in challenging real-world turbid underwater environments, and experimental results demonstrate that our method significantly outperforms existing approaches in terms of marker detection robustness and pose estimation accuracy. The code and dataset are publicly available at <span><span>https://github.com/fanyi-meng/DeepTurbid</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":50367,"journal":{"name":"Information Fusion","volume":"126 ","pages":"Article 103568"},"PeriodicalIF":15.5,"publicationDate":"2025-07-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144721905","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Rank-aware LDL hybrid MetaFormer for Compound Facial Expression Recognition in-the-wild 用于野外复合面部表情识别的等级感知LDL混合元former

IF 14.7 1区计算机科学

Information Fusion Pub Date : 2025-07-23 DOI: 10.1016/j.inffus.2025.103525

Afifa Khelifa , Haythem Ghazouani , Walid Barhoumi

{"title":"Rank-aware LDL hybrid MetaFormer for Compound Facial Expression Recognition in-the-wild","authors":"Afifa Khelifa , Haythem Ghazouani , Walid Barhoumi","doi":"10.1016/j.inffus.2025.103525","DOIUrl":"10.1016/j.inffus.2025.103525","url":null,"abstract":"<div><div>This study presents an effective deep learning-based method for Compound Facial Expression Recognition (CFER) in unconstrained environments. Conventional FER methods struggle with real-world complexity, where emotions often appear as nuanced combinations of basic expressions. To meet this challenge, we propose a method that takes advantage of label distribution learning within a hybrid MetaFormer architecture, which merges the strengths of CNNs and transformers, for better management of ambiguity and complexity. Moreover, our end-to-end method, called Top Rank Label Distribution Learning for Compound Facial Expression Recognition (TR-LDL-CFER), exploits a new combined loss function that learns both the label distribution and the ranking relationships between predicted labels, thus preserving crucial information for accurately inferring compound emotions. We validate the effectiveness of our method through extensive experiments on three benchmark in-the-wild CFER datasets and obtain competitive results with UARs of 41.42%, 50.77% and 51.36% on RAF-CE, RAF-DB Compound and EmotioNet, respectively.</div></div>","PeriodicalId":50367,"journal":{"name":"Information Fusion","volume":"126 ","pages":"Article 103525"},"PeriodicalIF":14.7,"publicationDate":"2025-07-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144702802","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Mitigating malicious model fusion in federated learning via confidence-aware defense 基于信心感知防御的联邦学习中的恶意模型融合

IF 14.7 1区计算机科学

Information Fusion Pub Date : 2025-07-23 DOI: 10.1016/j.inffus.2025.103529

Qilei Li , Pantelis Papageorgiou , Gaoyang Liu , Mingliang Gao , Linlin You , Chen Wang , Ahmed M. Abdelmoniem

{"title":"Mitigating malicious model fusion in federated learning via confidence-aware defense","authors":"Qilei Li , Pantelis Papageorgiou , Gaoyang Liu , Mingliang Gao , Linlin You , Chen Wang , Ahmed M. Abdelmoniem","doi":"10.1016/j.inffus.2025.103529","DOIUrl":"10.1016/j.inffus.2025.103529","url":null,"abstract":"<div><div>Federated Learning (FL) has emerged as a promising approach for decentralized machine learning in Internet of Things (IoT) applications, where privacy-sensitive data remains distributed across devices. However, FL systems are vulnerable to attacks that are happening in malicious clients via data poisoning and model poisoning. Once such malicious models are fused in the global server, it will deteriorate the global model’s performance. Existing defense methods typically mitigate specific types of poisoning but are often ineffective against others. To overcome this issue, we propose a simple yet effective framework called Confidence-Aware Defense (CAD). It aims to achieve accurate, robust, and versatile detection of malicious attacks. CAD evaluates the reliability of client updates by leveraging the confidence scores produced by each FL client model. Our key insight is that poisoning attacks, regardless of attack type, will cause the model to deviate from its previous state, thus leading to increased uncertainty when making predictions. Therefore, CAD is comprehensively effective for various types of poisoning attacks, including model poisoning and data poisoning. The proposed CAD method accurately identifies and mitigates malicious updates, even under varying attack intensities and data heterogeneity. CAD is evaluated on standard FL benchmarks (CIFAR-10, MNIST, Fashion-MNIST) under non-IID settings and both model and data poisoning attacks. It achieves up to 97.8% accuracy on MNIST and sustains over 64% accuracy under 50% poisoning on CIFAR-10. CAD surpasses all prior defense methods in robustness and performance. These results demonstrate the practicality of CAD in securing FL systems against various threat scenarios.</div></div>","PeriodicalId":50367,"journal":{"name":"Information Fusion","volume":"126 ","pages":"Article 103529"},"PeriodicalIF":14.7,"publicationDate":"2025-07-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144712890","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Quantum-inspired complex spiking neural network for multi-source data joint classification 多源数据联合分类的量子启发复杂脉冲神经网络

IF 15.5 1区计算机科学

Information Fusion Pub Date : 2025-07-22 DOI: 10.1016/j.inffus.2025.103526

Yang Liu, Yahui Li, Haige Xu, Yinghao Lin

{"title":"Quantum-inspired complex spiking neural network for multi-source data joint classification","authors":"Yang Liu, Yahui Li, Haige Xu, Yinghao Lin","doi":"10.1016/j.inffus.2025.103526","DOIUrl":"10.1016/j.inffus.2025.103526","url":null,"abstract":"<div><div>The joint classification of hyperspectral images (HSI) and Light Detection and Ranging (LiDAR) data shows great potential by leveraging the complementary characteristics of spectral information and elevation data. However, existing methods are difficult to achieve multi-source data fusion in the encoding stage, and generally have problems such as high architecture redundancy and high computing energy consumption, which are difficult to meet the application deployment of unmanned autonomous equipment such as satellite-borne and airborne. This paper proposes a quantum-inspired complex spiking neural network (QI-CSNN) framework to achieve efficient fusion and classification of multi-source data. Firstly, based on the characteristics of quantum superposition states, a complex spiking multi-source coding method is designed to achieve preliminary fusion of HSI and LiDAR data in the coding stage. Secondly, a complex spiking feature extraction module (CSFEM) is constructed based on complex convolution, complex batch normalization (BN), and leaky integrate-and-fire (LIF) neurons, which enhances the fusion and interaction of heterogeneous information through complex operations. Thirdly, a complex spiking feature extraction network is constructed through layer-by-layer concatenation of CSFEM, and multi-source data joint classification was achieved through probability amplitude. Finally, based on the implementation mechanism of QI-CSNN, the theoretical framework of multivariate neural network processing multi-source data classification is derived. Experimental results show that the proposed method achieves advanced classification accuracy. QI-CSNN provides a new solution for multi-source data joint classification tasks, which is of great significance for the deployment of multi-source data joint classification algorithms in unmanned autonomous equipment.</div></div>","PeriodicalId":50367,"journal":{"name":"Information Fusion","volume":"126 ","pages":"Article 103526"},"PeriodicalIF":15.5,"publicationDate":"2025-07-22","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144722052","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":1,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0