Information Fusion最新文献_第10页

TPIN: Text-based parallel interaction network with modality-common and modality-specific for multimodal sentiment analysis 基于多情态情感分析的情态通用和情态特定的文本并行交互网络

IF 15.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Information Fusion

Pub Date : 2026-06-01 Epub Date: 2025-12-24 DOI: 10.1016/j.inffus.2025.104087

Changbin Wang, Fengrui Ji, Baolin Liu

Learning an effective joint representation is fundamental for Multimodal Sentiment Analysis (MSA). Existing studies typically adopt complex networks to construct joint multimodal representations directly, yet often overlook the heterogeneity among different modalities as well as the preservation of modality-specific information. Moreover, current methods tend to treat all modalities equally, failing to exploit the rich emotional cues in the text modality. To address these issues, we propose a Text-based Parallel Interaction Network (TPIN) that aims to trade off the commonality and specificity of different modalities. The TPIN consists of two components: Modality-Common Information Processing (MCIP) and Modality-Specific Information Processing (MSIP). In MCIP, we innovatively propose a contrastive learning algorithm with Hard Negative Mining (HNM), which is integrated into our designed Two-Stage Contrastive Learning (TSCL) to mitigate inter-modal heterogeneity. Additionally, we design a Text-Guided Dynamic Semantic Aggregation (TG-DSA) module to enable deep multimodal fusion under the guidance of text modality. In MSIP, we devise a dynamic routing mechanism, which iteratively optimizes routing weights to better capture modality-specific information in visual and acoustic modalities. Experimental results demonstrate that our method achieves state-of-the-art performance on both the CMU-MOSI and CMU-MOSEI datasets, showing consistent gains of 0.5%–1.2% across major evaluation metrics compared with recent advanced models.

学习有效的联合表示是多模态情感分析（MSA）的基础。现有研究通常采用复杂网络直接构建联合多模态表征，但往往忽略了不同模态之间的异质性以及模态特有信息的保存。此外，现有的方法倾向于平等对待所有的情态，未能充分利用语篇情态中丰富的情感线索。为了解决这些问题，我们提出了一个基于文本的并行交互网络（TPIN），旨在权衡不同模式的共性和特殊性。TPIN由两部分组成：模态通用信息处理（MCIP）和模态特定信息处理（MSIP）。在MCIP中，我们创新地提出了一种带有硬负挖掘（HNM）的对比学习算法，并将其集成到我们设计的两阶段对比学习（TSCL）中，以减轻模态间的异质性。此外，我们设计了文本引导动态语义聚合（TG-DSA）模块，在文本模态的引导下实现深度多模态融合。在MSIP中，我们设计了一个动态路由机制，该机制迭代优化路由权重，以更好地捕获视觉和声学模态中的特定模态信息。实验结果表明，我们的方法在CMU-MOSI和CMU-MOSEI数据集上都达到了最先进的性能，与最近的先进模型相比，在主要评估指标上显示出0.5%-1.2%的一致增益。

{"title":"TPIN: Text-based parallel interaction network with modality-common and modality-specific for multimodal sentiment analysis","authors":"Changbin Wang, Fengrui Ji, Baolin Liu","doi":"10.1016/j.inffus.2025.104087","DOIUrl":"10.1016/j.inffus.2025.104087","url":null,"abstract":"<div><div>Learning an effective joint representation is fundamental for Multimodal Sentiment Analysis (MSA). Existing studies typically adopt complex networks to construct joint multimodal representations directly, yet often overlook the heterogeneity among different modalities as well as the preservation of modality-specific information. Moreover, current methods tend to treat all modalities equally, failing to exploit the rich emotional cues in the text modality. To address these issues, we propose a Text-based Parallel Interaction Network (TPIN) that aims to trade off the commonality and specificity of different modalities. The TPIN consists of two components: Modality-Common Information Processing (MCIP) and Modality-Specific Information Processing (MSIP). In MCIP, we innovatively propose a contrastive learning algorithm with Hard Negative Mining (HNM), which is integrated into our designed Two-Stage Contrastive Learning (TSCL) to mitigate inter-modal heterogeneity. Additionally, we design a Text-Guided Dynamic Semantic Aggregation (TG-DSA) module to enable deep multimodal fusion under the guidance of text modality. In MSIP, we devise a dynamic routing mechanism, which iteratively optimizes routing weights to better capture modality-specific information in visual and acoustic modalities. Experimental results demonstrate that our method achieves state-of-the-art performance on both the CMU-MOSI and CMU-MOSEI datasets, showing consistent gains of 0.5%–1.2% across major evaluation metrics compared with recent advanced models.</div></div>","PeriodicalId":50367,"journal":{"name":"Information Fusion","volume":"130 ","pages":"Article 104087"},"PeriodicalIF":15.5,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145822967","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

Rethink: reveal the impact of semantic distribution transfer from the cross-modal hashing perspective 重新思考：从跨模态散列的角度揭示语义分布转移的影响

IF 15.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Information Fusion

Pub Date : 2026-06-01 Epub Date: 2026-01-05 DOI: 10.1016/j.inffus.2026.104123

Yinan Li , Zhi Liu , Jiajun Tang , Binghong Chen , Mingjin Kuai , Jun Long , Zhan Yang

Hashing has been extensively applied in cross-modal retrieval by mapping diverse modalities data into binary codes. Semantic transfer aims to enhance the relevance of heterogeneous representations through migrating valuable information from one modality to another in the unsupervised paradigm. The combination of semantic transfer and hash learning substitutes the dense vector search with Hamming distance, significantly reducing storage requirements and increasing retrieval efficiency. However, the current unsupervised mechanism demonstrates ordinary performance in retrieval precision, which requires more improvement from semantic annotation. Particularly, the mediocre information fusion strategy directly affects the quality of learned hash codes. In this paper, we propose a novel Semantic Transfer framework for Semi-supervised Cross-modal Hashing, denoted as STSCH. Initially, we utilize multiple auto-encoders to learn the high-level semantic representation of each modality. To guarantee the completeness of heterogeneous data, we incorporate them via semantic transfer and analyse the feature distribution of diverse modalities. Furthermore, an asymmetric hash learning framework between individual modality-specific representation and minor semantic labels is constructed. Finally, an effective optimization algorithm is proposed. Comprehensive experiments on Wiki, MIRFlickr, and NUS-WIDE datasets demonstrate the superior performance of STSCH to state-of-the-art hashing approaches.

通过将不同模态的数据映射成二进制码，哈希在跨模态检索中得到了广泛应用。语义迁移旨在通过在无监督范式中将有价值的信息从一种模态迁移到另一种模态来增强异构表示的相关性。语义转移与哈希学习的结合替代了基于汉明距离的密集向量搜索，显著降低了存储需求，提高了检索效率。然而，目前的无监督机制在检索精度上表现一般，还需要语义标注的进一步改进。其中，信息融合策略的平庸性直接影响了学习到的哈希码的质量。在本文中，我们提出了一种新的半监督跨模态哈希语义转移框架，称为STSCH。首先，我们使用多个自编码器来学习每个模态的高级语义表示。为了保证异构数据的完整性，我们通过语义转移对异构数据进行整合，并分析了不同模态的特征分布。此外，在单个模态特定表示和次要语义标签之间构建了一个非对称哈希学习框架。最后，提出了一种有效的优化算法。在Wiki、MIRFlickr和NUS-WIDE数据集上的综合实验表明，STSCH比最先进的哈希方法性能优越。

{"title":"Rethink: reveal the impact of semantic distribution transfer from the cross-modal hashing perspective","authors":"Yinan Li , Zhi Liu , Jiajun Tang , Binghong Chen , Mingjin Kuai , Jun Long , Zhan Yang","doi":"10.1016/j.inffus.2026.104123","DOIUrl":"10.1016/j.inffus.2026.104123","url":null,"abstract":"<div><div>Hashing has been extensively applied in cross-modal retrieval by mapping diverse modalities data into binary codes. Semantic transfer aims to enhance the relevance of heterogeneous representations through migrating valuable information from one modality to another in the unsupervised paradigm. The combination of semantic transfer and hash learning substitutes the dense vector search with Hamming distance, significantly reducing storage requirements and increasing retrieval efficiency. However, the current unsupervised mechanism demonstrates ordinary performance in retrieval precision, which requires more improvement from semantic annotation. Particularly, the mediocre information fusion strategy directly affects the quality of learned hash codes. In this paper, we propose a novel <strong>S</strong>emantic <strong>T</strong>ransfer framework for <strong>S</strong>emi-supervised <strong>C</strong>ross-modal <strong>H</strong>ashing, denoted as STSCH. Initially, we utilize multiple auto-encoders to learn the high-level semantic representation of each modality. To guarantee the completeness of heterogeneous data, we incorporate them via semantic transfer and analyse the feature distribution of diverse modalities. Furthermore, an asymmetric hash learning framework between individual modality-specific representation and minor semantic labels is constructed. Finally, an effective optimization algorithm is proposed. Comprehensive experiments on Wiki, MIRFlickr, and NUS-WIDE datasets demonstrate the superior performance of STSCH to state-of-the-art hashing approaches.</div></div>","PeriodicalId":50367,"journal":{"name":"Information Fusion","volume":"130 ","pages":"Article 104123"},"PeriodicalIF":15.5,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145902475","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

Multimodal brain network analysis: Research advances and challenges 多模态脑网络分析：研究进展与挑战

IF 15.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Information Fusion

Pub Date : 2026-06-01 Epub Date: 2025-12-26 DOI: 10.1016/j.inffus.2025.104096

Tao Hou , Wenhao Dai , Jiashuang Huang , Youyong Kong , Weiping Ding

Brain network analysis has become a powerful approach in neuroscience for characterizing the structural and functional organization of brain regions. Advances in neuroimaging techniques, including structural magnetic resonance imaging (sMRI), functional magnetic resonance imaging (fMRI), diffusion tensor imaging (DTI), and electroencephalography (EEG), provide complementary information on brain anatomy, connectivity patterns, and electrophysiological dynamics across different temporal and spatial scales. While unimodal analyses provide valuable insights, they are inherently limited in capturing the complexity and dynamics of brain networks. Multimodal fusion strategies have therefore become crucial for constructing more accurate and informative brain network models, enabling applications such as neurological disorder classification, brain age prediction, and cognitive function assessment. In this review, we systematically survey contemporary deep learning architectures for brain network analysis, including traditional frameworks and emerging technologies such as large language models. We further review multimodal fusion strategies, highlighting recent methodological advances, their relevance to clinical and cognitive applications, and the role of interpretability in enhancing model understanding and revealing neural mechanisms. We also discuss open technical challenges and consider potential directions for future research, including strategies to address individual variability and to ensure data privacy and security in multimodal brain network modeling. This review provides a comprehensive and forward-looking reference for researchers seeking to advance the understanding and practical application of complex brain networks.

脑网络分析已成为表征脑区域结构和功能组织的一种强有力的神经科学方法。包括结构磁共振成像（sMRI）、功能磁共振成像（fMRI）、弥散张量成像（DTI）和脑电图（EEG）在内的神经成像技术的进步，为不同时空尺度的大脑解剖、连接模式和电生理动力学提供了补充信息。虽然单峰分析提供了有价值的见解，但它们在捕捉大脑网络的复杂性和动态性方面存在固有的局限性。因此，多模态融合策略对于构建更准确和信息丰富的脑网络模型至关重要，从而实现神经系统疾病分类、脑年龄预测和认知功能评估等应用。在这篇综述中，我们系统地调查了当代用于大脑网络分析的深度学习架构，包括传统框架和新兴技术，如大型语言模型。我们进一步回顾了多模态融合策略，强调了最近的方法进展，它们与临床和认知应用的相关性，以及可解释性在增强模型理解和揭示神经机制方面的作用。我们还讨论了开放的技术挑战，并考虑了未来研究的潜在方向，包括解决个体可变性的策略，以及确保多模态大脑网络建模中的数据隐私和安全。这一综述为寻求推进复杂脑网络的理解和实际应用的研究人员提供了全面和前瞻性的参考。

{"title":"Multimodal brain network analysis: Research advances and challenges","authors":"Tao Hou , Wenhao Dai , Jiashuang Huang , Youyong Kong , Weiping Ding","doi":"10.1016/j.inffus.2025.104096","DOIUrl":"10.1016/j.inffus.2025.104096","url":null,"abstract":"<div><div>Brain network analysis has become a powerful approach in neuroscience for characterizing the structural and functional organization of brain regions. Advances in neuroimaging techniques, including structural magnetic resonance imaging (sMRI), functional magnetic resonance imaging (fMRI), diffusion tensor imaging (DTI), and electroencephalography (EEG), provide complementary information on brain anatomy, connectivity patterns, and electrophysiological dynamics across different temporal and spatial scales. While unimodal analyses provide valuable insights, they are inherently limited in capturing the complexity and dynamics of brain networks. Multimodal fusion strategies have therefore become crucial for constructing more accurate and informative brain network models, enabling applications such as neurological disorder classification, brain age prediction, and cognitive function assessment. In this review, we systematically survey contemporary deep learning architectures for brain network analysis, including traditional frameworks and emerging technologies such as large language models. We further review multimodal fusion strategies, highlighting recent methodological advances, their relevance to clinical and cognitive applications, and the role of interpretability in enhancing model understanding and revealing neural mechanisms. We also discuss open technical challenges and consider potential directions for future research, including strategies to address individual variability and to ensure data privacy and security in multimodal brain network modeling. This review provides a comprehensive and forward-looking reference for researchers seeking to advance the understanding and practical application of complex brain networks.</div></div>","PeriodicalId":50367,"journal":{"name":"Information Fusion","volume":"130 ","pages":"Article 104096"},"PeriodicalIF":15.5,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145845114","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

SynJAC: synthetic-data-driven joint-granular adaptation and calibration for domain specific scanned document key information extraction 面向特定领域扫描文档关键信息提取的综合数据驱动的关节颗粒自适应与校准

IF 15.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Information Fusion

Pub Date : 2026-06-01 Epub Date: 2025-12-23 DOI: 10.1016/j.inffus.2025.104074

Yihao Ding , Soyeon Caren Han , Zechuan Li , Hyunsuk Chung

Visually Rich Documents (VRDs), comprising elements such as charts, tables, and paragraphs, convey complex information across diverse domains. However, extracting key information from these documents remains labour-intensive, particularly for scanned formats with inconsistent layouts and domain-specific requirements. Despite advances in pretrained models for VRD understanding, their dependence on large annotated datasets for fine-tuning hinders scalability. This paper proposes SynJAC (Synthetic-data-driven Joint-granular Adaptation and Calibration), a method for key information extraction in scanned documents. SynJAC leverages synthetic, machine-generated data for domain adaptation and employs calibration on a small, manually annotated dataset to mitigate noise. By integrating fine-grained and coarse-grained document representation learning, SynJAC significantly reduces the need for extensive manual labelling while achieving competitive performance. Extensive experiments demonstrate its effectiveness in domain-specific and scanned VRD scenarios.

视觉上丰富的文档（vrd）由图表、表格和段落等元素组成，在不同的领域传递复杂的信息。然而，从这些文档中提取关键信息仍然是劳动密集型的，特别是对于具有不一致布局和特定领域要求的扫描格式。尽管在VRD理解的预训练模型方面取得了进展，但它们依赖于大型带注释的数据集进行微调，阻碍了可扩展性。提出了一种基于SynJAC （Synthetic-data-driven Joint-granular Adaptation and Calibration）的扫描文档关键信息提取方法。SynJAC利用合成的、机器生成的数据进行域适应，并在一个小的、手动注释的数据集上进行校准，以减轻噪声。通过集成细粒度和粗粒度文档表示学习，SynJAC显著减少了大量手工标记的需要，同时实现了具有竞争力的性能。大量的实验证明了该方法在特定领域和扫描VRD场景下的有效性。

{"title":"SynJAC: synthetic-data-driven joint-granular adaptation and calibration for domain specific scanned document key information extraction","authors":"Yihao Ding , Soyeon Caren Han , Zechuan Li , Hyunsuk Chung","doi":"10.1016/j.inffus.2025.104074","DOIUrl":"10.1016/j.inffus.2025.104074","url":null,"abstract":"<div><div>Visually Rich Documents (VRDs), comprising elements such as charts, tables, and paragraphs, convey complex information across diverse domains. However, extracting key information from these documents remains labour-intensive, particularly for scanned formats with inconsistent layouts and domain-specific requirements. Despite advances in pretrained models for VRD understanding, their dependence on large annotated datasets for fine-tuning hinders scalability. This paper proposes <strong>SynJAC</strong> (Synthetic-data-driven Joint-granular Adaptation and Calibration), a method for key information extraction in scanned documents. SynJAC leverages synthetic, machine-generated data for domain adaptation and employs calibration on a small, manually annotated dataset to mitigate noise. By integrating fine-grained and coarse-grained document representation learning, SynJAC significantly reduces the need for extensive manual labelling while achieving competitive performance. Extensive experiments demonstrate its effectiveness in domain-specific and scanned VRD scenarios.</div></div>","PeriodicalId":50367,"journal":{"name":"Information Fusion","volume":"130 ","pages":"Article 104074"},"PeriodicalIF":15.5,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145822974","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

ViP-HMNN: a visual pathway-inspired hybrid neural network incorporated with in-memory computing for object recognition ViP-HMNN：一种结合内存计算的视觉路径启发的混合神经网络，用于对象识别

IF 15.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Information Fusion

Pub Date : 2026-06-01 Epub Date: 2025-12-22 DOI: 10.1016/j.inffus.2025.104086

Xiaoyue Ji , Yifeng Han , Chun Sing Lai , Fangwen Yu , Zhekang Dong

The integration of artificial neural networks (ANNs) and spiking neural networks (SNNs) has significant potential for advancing artificial general intelligence (AGI). However, the hardware design of hybrid neural networks (HNNs) still primarily relies on near-memory computing architectures, which have yet to fully overcome the separation between processing and memory units. To address this, we develop a visual pathway-inspired hybrid memristive neural network (ViP-HMNN). Specifically, we design a general and compact memristor-based neuron circuit that can efficiently implement both ANN and SNN activation functions, serving as the core component of the proposed ViP-HMNN. To improve the understanding of the designed ViP-HMNN, a ventral pathway-inspired static feature extraction module (VP-SFEM), a dorsal pathway-inspired dynamic feature representation module (DP-DFRM), and a complementary feature fusion output module (CFFOM) are proposed. For validation, the proposed ViP-HMNN with synergistic hybrid training strategy is applied for object recognition. Compared to software-based object recognition methods, the proposed ViP-HMNN achieves software-compatible accuracy (ranking in the top three) and has a significant advantage in terms of time consumption (at least 8-fold faster). Compared to in-memory computing architectures, the proposed ViP-HMNN exhibits enhancements of up to 2.65× in area overhead, 1.83× in latency, and 1.87× in energy consumption. Compared to the RTX 3090 GPU, the proposed ViP-HMNN achieves a minimum of 1000× reduction in latency and 500× in energy saving.

人工神经网络（ANNs）和脉冲神经网络（SNNs）的融合在推进人工通用智能（AGI）方面具有巨大的潜力。然而，混合神经网络（HNNs）的硬件设计仍然主要依赖于近内存计算架构，这还没有完全克服处理和存储单元之间的分离。为了解决这个问题，我们开发了一种视觉通路启发的混合记忆神经网络（ViP-HMNN）。具体来说，我们设计了一个通用的、紧凑的基于忆阻器的神经元电路，可以有效地实现ANN和SNN的激活函数，作为所提出的ViP-HMNN的核心组件。为了提高对设计的ViP-HMNN的理解，提出了腹侧路径启发的静态特征提取模块（VP-SFEM）、背侧路径启发的动态特征表示模块（DP-DFRM）和互补特征融合输出模块（CFFOM）。为了验证所提出的具有协同混合训练策略的ViP-HMNN在目标识别中的应用。与基于软件的目标识别方法相比，本文提出的ViP-HMNN实现了软件兼容的精度（排名前三），并且在时间消耗方面具有显著优势（至少快8倍）。与内存计算体系结构相比，所提出的vip - hmmn在面积开销、延迟和能耗方面分别提高了2.65倍、1.83倍和1.87倍。与RTX 3090 GPU相比，本文提出的vip - hmmn至少减少了1000倍的延迟，节省了500倍的能源。

{"title":"ViP-HMNN: a visual pathway-inspired hybrid neural network incorporated with in-memory computing for object recognition","authors":"Xiaoyue Ji , Yifeng Han , Chun Sing Lai , Fangwen Yu , Zhekang Dong","doi":"10.1016/j.inffus.2025.104086","DOIUrl":"10.1016/j.inffus.2025.104086","url":null,"abstract":"<div><div>The integration of artificial neural networks (ANNs) and spiking neural networks (SNNs) has significant potential for advancing artificial general intelligence (AGI). However, the hardware design of hybrid neural networks (HNNs) still primarily relies on near-memory computing architectures, which have yet to fully overcome the separation between processing and memory units. To address this, we develop a visual pathway-inspired hybrid memristive neural network (ViP-HMNN). Specifically, we design a general and compact memristor-based neuron circuit that can efficiently implement both ANN and SNN activation functions, serving as the core component of the proposed ViP-HMNN. To improve the understanding of the designed ViP-HMNN, a ventral pathway-inspired static feature extraction module (VP-SFEM), a dorsal pathway-inspired dynamic feature representation module (DP-DFRM), and a complementary feature fusion output module (CFFOM) are proposed. For validation, the proposed ViP-HMNN with synergistic hybrid training strategy is applied for object recognition. Compared to software-based object recognition methods, the proposed ViP-HMNN achieves software-compatible accuracy (ranking in the top three) and has a significant advantage in terms of time consumption (at least 8-fold faster). Compared to in-memory computing architectures, the proposed ViP-HMNN exhibits enhancements of up to 2.65× in area overhead, 1.83× in latency, and 1.87× in energy consumption. Compared to the RTX 3090 GPU, the proposed ViP-HMNN achieves a minimum of 1000× reduction in latency and 500× in energy saving.</div></div>","PeriodicalId":50367,"journal":{"name":"Information Fusion","volume":"130 ","pages":"Article 104086"},"PeriodicalIF":15.5,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145808543","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

MoMD Transformer: adaptive multi-modal fault diagnosis via knowledge transfer with vibration-current signals MoMD变压器：基于振动电流信号知识传递的自适应多模态故障诊断

IF 15.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Information Fusion

Pub Date : 2026-06-01 Epub Date: 2025-12-22 DOI: 10.1016/j.inffus.2025.104079

Xiaohan Zhang , Han Wang , Chenze Wang , Gaowei Xu , Min Liu , Chuang Liu

Fault diagnosis is crucial to ensure the reliable and safe operation of electromechanical systems in the manufacturing industry. Recently, multi-modal data fusion based fault diagnosis methods have achieved remarkable progress. However, these methods impose additional constraints on industrial scenarios because they require the simultaneous input of multi-modal data during the inference phase, which limits their applicability in cases where only single-modal data acquisition is feasible during operation. To overcome this limitation, a novel Mixture-of-Modality-Diagnosis (MoMD) Transformer is proposed in this paper for adaptive multi-modal fault diagnosis with vibration and current signals. In this model, we improve the feed-forward network in Transformer into a multi-channel structure to adaptively process multiple modalities. Additionally, to address the challenge of weak fault signatures in current modality, we design a global knowledge transfer module that utilizes vibration features to guide the feature learning for current signals. Specifically, our comprehensive analysis of feature alignment strategies reveals that the one-to-one paradigm is superior to contrastive learning paradigm for fault data, given its low intra-class variance. Moreover, to enhance the representation ability for time-series signals, the masked signal modeling task is introduced into the training phase, thereby improving the diagnostic accuracy. Experimental results on two datasets demonstrate that the proposed method achieves superior diagnostic accuracy, reaching 99.96 % and 100 % respectively, outperforming other methods in various modality-availability scenarios.

在制造业中，故障诊断是保证机电系统可靠、安全运行的关键。近年来，基于多模态数据融合的故障诊断方法取得了显著进展。然而，这些方法对工业场景施加了额外的约束，因为它们需要在推理阶段同时输入多模态数据，这限制了它们在操作过程中只有单模态数据采集可行的情况下的适用性。为了克服这一局限性，本文提出了一种新的混合模态诊断变压器，用于振动和电流信号的自适应多模态故障诊断。在该模型中，我们将Transformer中的前馈网络改进为多通道结构，以自适应处理多种模态。此外，为了解决当前模态中弱故障特征的挑战，我们设计了一个利用振动特征指导当前信号特征学习的全局知识传递模块。具体地说，我们对特征对齐策略的综合分析表明，一对一范式优于对比学习范式，因为它具有较低的类内方差。此外，为了增强对时间序列信号的表示能力，在训练阶段引入了屏蔽信号建模任务，从而提高了诊断准确率。在两个数据集上的实验结果表明，该方法的诊断准确率分别达到99.96%和100%，在各种模态可用性场景下均优于其他方法。

{"title":"MoMD Transformer: adaptive multi-modal fault diagnosis via knowledge transfer with vibration-current signals","authors":"Xiaohan Zhang , Han Wang , Chenze Wang , Gaowei Xu , Min Liu , Chuang Liu","doi":"10.1016/j.inffus.2025.104079","DOIUrl":"10.1016/j.inffus.2025.104079","url":null,"abstract":"<div><div>Fault diagnosis is crucial to ensure the reliable and safe operation of electromechanical systems in the manufacturing industry. Recently, multi-modal data fusion based fault diagnosis methods have achieved remarkable progress. However, these methods impose additional constraints on industrial scenarios because they require the simultaneous input of multi-modal data during the inference phase, which limits their applicability in cases where only single-modal data acquisition is feasible during operation. To overcome this limitation, a novel Mixture-of-Modality-Diagnosis (MoMD) Transformer is proposed in this paper for adaptive multi-modal fault diagnosis with vibration and current signals. In this model, we improve the feed-forward network in Transformer into a multi-channel structure to adaptively process multiple modalities. Additionally, to address the challenge of weak fault signatures in current modality, we design a global knowledge transfer module that utilizes vibration features to guide the feature learning for current signals. Specifically, our comprehensive analysis of feature alignment strategies reveals that the one-to-one paradigm is superior to contrastive learning paradigm for fault data, given its low intra-class variance. Moreover, to enhance the representation ability for time-series signals, the masked signal modeling task is introduced into the training phase, thereby improving the diagnostic accuracy. Experimental results on two datasets demonstrate that the proposed method achieves superior diagnostic accuracy, reaching 99.96 % and 100 % respectively, outperforming other methods in various modality-availability scenarios.</div></div>","PeriodicalId":50367,"journal":{"name":"Information Fusion","volume":"130 ","pages":"Article 104079"},"PeriodicalIF":15.5,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145823826","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

WDASR: A wavelet-based deformable attention network for cardiac cine MRI super-resolution with spatiotemporal motion modeling WDASR：一种基于小波的可变形注意网络，用于心脏电影MRI超分辨率的时空运动建模

IF 15.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Information Fusion

Pub Date : 2026-06-01 Epub Date: 2026-01-06 DOI: 10.1016/j.inffus.2025.104116

Jun Lyu , Xunkang Zhao , Jing Qin , Chengyan Wang

Cardiac cine MRI is the clinical gold standard for dynamic cardiac assessment, but reducing k-space sampling to accelerate acquisition results in low-resolution images that fail to depict fine anatomical details. Existing super-resolution methods struggle to preserve spatial details and temporal coherence due to limitations in handling non-rigid cardiac deformations and lossy feature downsampling. This paper proposes a Wavelet-based Deformable Attention Super-Resolution Network (WDASR) that addresses these limitations through two key innovations. First, a Frequency Subband Adaptive Alignment (FSAA) module applies deformable convolution to wavelet-decomposed frequency subbands, enabling lossless downsampling that prevents offset over-shifting and allows targeted alignment across neighboring and remote frames. Second, a Cross-Resolution Wavelet Attention (CRWA) module uses temporally-aggregated frequency subbands as low-resolution keys and values, and the current frame as high-resolution query, reducing computational complexity by 75% while effectively integrating multi-scale spatiotemporal information for enhanced texture representation. A bidirectional recurrent mechanism further propagates the enhanced features to maintain temporal consistency. Experiments on public and private datasets demonstrate that WDASR achieves 4 ×  super-resolution with state-of-the-art performance and potential for clinical application.

心脏电影MRI是动态心脏评估的临床金标准，但减少k空间采样以加速采集会导致低分辨率图像无法描绘精细的解剖细节。由于处理非刚性心脏变形和有损特征下采样的局限性，现有的超分辨率方法难以保持空间细节和时间相干性。本文提出了一种基于小波的可变形注意力超分辨率网络（WDASR），通过两个关键创新解决了这些限制。首先，频率子带自适应对准（FSAA）模块对小波分解的频率子带进行可变形卷积，实现无损下采样，防止偏移过移，并允许在相邻帧和远程帧之间进行目标对准。其次，交叉分辨率小波注意（Cross-Resolution Wavelet Attention， CRWA）模块采用时间聚合的频率子带作为低分辨率键和值，当前帧作为高分辨率查询，在有效整合多尺度时空信息的同时，将计算复杂度降低了75%，增强了纹理表征。双向循环机制进一步传播增强的特征以保持时间一致性。在公共和私有数据集上的实验表明，WDASR达到了4 × 超分辨率，具有最先进的性能和临床应用潜力。

{"title":"WDASR: A wavelet-based deformable attention network for cardiac cine MRI super-resolution with spatiotemporal motion modeling","authors":"Jun Lyu , Xunkang Zhao , Jing Qin , Chengyan Wang","doi":"10.1016/j.inffus.2025.104116","DOIUrl":"10.1016/j.inffus.2025.104116","url":null,"abstract":"<div><div>Cardiac cine MRI is the clinical gold standard for dynamic cardiac assessment, but reducing k-space sampling to accelerate acquisition results in low-resolution images that fail to depict fine anatomical details. Existing super-resolution methods struggle to preserve spatial details and temporal coherence due to limitations in handling non-rigid cardiac deformations and lossy feature downsampling. This paper proposes a Wavelet-based Deformable Attention Super-Resolution Network (WDASR) that addresses these limitations through two key innovations. First, a Frequency Subband Adaptive Alignment (FSAA) module applies deformable convolution to wavelet-decomposed frequency subbands, enabling lossless downsampling that prevents offset over-shifting and allows targeted alignment across neighboring and remote frames. Second, a Cross-Resolution Wavelet Attention (CRWA) module uses temporally-aggregated frequency subbands as low-resolution keys and values, and the current frame as high-resolution query, reducing computational complexity by 75% while effectively integrating multi-scale spatiotemporal information for enhanced texture representation. A bidirectional recurrent mechanism further propagates the enhanced features to maintain temporal consistency. Experiments on public and private datasets demonstrate that WDASR achieves 4 ×  super-resolution with state-of-the-art performance and potential for clinical application.</div></div>","PeriodicalId":50367,"journal":{"name":"Information Fusion","volume":"130 ","pages":"Article 104116"},"PeriodicalIF":15.5,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145978237","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

Multi-modal and multi-condition fault diagnosis of rotating machinery via a heterogeneous graph learning framework 基于异构图学习框架的旋转机械多模态多工况故障诊断

IF 15.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Information Fusion

Pub Date : 2026-06-01 Epub Date: 2025-12-29 DOI: 10.1016/j.inffus.2025.104106

Xiaoyu Han , Yunpeng Cao , Pan Hu , Weixing Feng

Intelligent fault diagnosis of rotating machinery under multi-modal and multi-condition scenarios presents critical challenges, including the low utilization of structural information and weak model generalization capability. To address these issues, this paper proposes a Structural Awareness Heterogeneous Graph Transformer (SAHGT) framework to achieve unified modeling and robust representation learning of multi-source monitoring signals. The method constructs a unified heterogeneous graph structure to integrate modal features, frequency-domain relationships, and spatial priors. By introducing a heterogeneous graph Transformer with a dual-guided attention mechanism that combines modality-guided and frequency-guided attention, it enhances the selective expression of key features and the discriminative capability for fault patterns. To enhance the model’s adaptability in non-stationary environments, an augmented view-driven contrastive learning mechanism is designed to further strengthen robustness against structural variations and distribution shifts. Notably, this paper establishes a unified training framework that enables switching between Domain Generalization (DG) and Domain Adaptation (DA) tasks solely by configuring loss function combinations without modifying the model architecture. The validation experiments conducted on a high-fidelity gas turbine test platform demonstrate the superior performance of the proposed SAHGT framework, achieving average fault diagnosis accuracies of 83.46% and 99.57% on nine DG tasks and twelve DA tasks, respectively. These results significantly outperform state-of-the-art graph neural network methods, highlighting the model’s strong cross-domain generalization and domain adaptability.

旋转机械在多模态、多工况下的智能故障诊断面临着结构信息利用率低、模型泛化能力弱等严峻挑战。为了解决这些问题，本文提出了一种结构感知异构图转换器（SAHGT）框架，以实现多源监测信号的统一建模和鲁棒表示学习。该方法构建了统一的异构图结构，集成了模态特征、频域关系和空间先验。通过引入具有模态引导和频率引导相结合的双引导注意机制的异构图转换器，增强了关键特征的选择性表达和故障模式的判别能力。为了增强模型在非平稳环境中的适应性，设计了一种增强的视图驱动对比学习机制，以进一步增强模型对结构变化和分布变化的鲁棒性。值得注意的是，本文建立了一个统一的训练框架，仅通过配置损失函数组合就可以在不修改模型体系结构的情况下在域泛化（DG）和域自适应（DA）任务之间切换。在高保真燃气轮机测试平台上进行的验证实验表明，所提出的SAHGT框架在9个DG任务和12个DA任务上的平均故障诊断准确率分别达到83.46%和99.57%。这些结果明显优于最先进的图神经网络方法，突出了模型强大的跨域泛化和域适应性。

{"title":"Multi-modal and multi-condition fault diagnosis of rotating machinery via a heterogeneous graph learning framework","authors":"Xiaoyu Han , Yunpeng Cao , Pan Hu , Weixing Feng","doi":"10.1016/j.inffus.2025.104106","DOIUrl":"10.1016/j.inffus.2025.104106","url":null,"abstract":"<div><div>Intelligent fault diagnosis of rotating machinery under multi-modal and multi-condition scenarios presents critical challenges, including the low utilization of structural information and weak model generalization capability. To address these issues, this paper proposes a Structural Awareness Heterogeneous Graph Transformer (SAHGT) framework to achieve unified modeling and robust representation learning of multi-source monitoring signals. The method constructs a unified heterogeneous graph structure to integrate modal features, frequency-domain relationships, and spatial priors. By introducing a heterogeneous graph Transformer with a dual-guided attention mechanism that combines modality-guided and frequency-guided attention, it enhances the selective expression of key features and the discriminative capability for fault patterns. To enhance the model’s adaptability in non-stationary environments, an augmented view-driven contrastive learning mechanism is designed to further strengthen robustness against structural variations and distribution shifts. Notably, this paper establishes a unified training framework that enables switching between Domain Generalization (DG) and Domain Adaptation (DA) tasks solely by configuring loss function combinations without modifying the model architecture. The validation experiments conducted on a high-fidelity gas turbine test platform demonstrate the superior performance of the proposed SAHGT framework, achieving average fault diagnosis accuracies of 83.46% and 99.57% on nine DG tasks and twelve DA tasks, respectively. These results significantly outperform state-of-the-art graph neural network methods, highlighting the model’s strong cross-domain generalization and domain adaptability.</div></div>","PeriodicalId":50367,"journal":{"name":"Information Fusion","volume":"130 ","pages":"Article 104106"},"PeriodicalIF":15.5,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145939776","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

Data science: a natural ecosystem 数据科学：一个自然生态系统

IF 15.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Information Fusion

Pub Date : 2026-06-01 Epub Date: 2026-01-12 DOI: 10.1016/j.inffus.2025.104113

Emilio Porcu , Roy El Moukari , Laurent Najman , Francisco Herrera , Horst Simon

This manuscript provides a systemic and data-centric view of what we term essential data science, as a natural ecosystem with challenges and missions stemming from the fusion of data universe with its multiple combinations of the 5D complexities (data structure, domain, cardinality, causality, and ethics) with the phases of the data life cycle. Data agents perform tasks driven by specific goals. The data scientist is an abstract entity that comes from the logical organization of data agents with their actions. Data scientists face challenges that are defined according to the missions. We define specific discipline-induced data science, which in turn allows for the definition of pan-data science, a natural ecosystem that integrates specific disciplines with the essential data science. We semantically split the essential data science into computational, and foundational. By formalizing this ecosystemic view, we contribute a general-purpose, fusion-oriented architecture for integrating heterogeneous knowledge, agents, and workflows-relevant to a wide range of disciplines and high-impact applications.

本文提供了一个系统的和以数据为中心的观点，我们称之为基本数据科学，作为一个自然生态系统，其挑战和任务源于数据宇宙与5D复杂性（数据结构、领域、基数、因果关系和伦理）的多种组合与数据生命周期阶段的融合。数据代理执行由特定目标驱动的任务。数据科学家是一个抽象的实体，它来自数据代理及其操作的逻辑组织。数据科学家面临的挑战是根据任务定义的。我们定义了特定学科诱导的数据科学，这反过来又允许定义泛数据科学，这是一个将特定学科与基本数据科学集成在一起的自然生态系统。我们从语义上将基本数据科学分为计算型和基础型。通过形式化这个生态系统视图，我们提供了一个通用的、面向融合的体系结构，用于集成异构知识、代理和工作流程——与广泛的学科和高影响力的应用相关。

{"title":"Data science: a natural ecosystem","authors":"Emilio Porcu , Roy El Moukari , Laurent Najman , Francisco Herrera , Horst Simon","doi":"10.1016/j.inffus.2025.104113","DOIUrl":"10.1016/j.inffus.2025.104113","url":null,"abstract":"<div><div>This manuscript provides a systemic and data-centric view of what we term <em>essential</em> data science, as a <em>natural</em> ecosystem with challenges and missions stemming from the fusion of data universe with its multiple combinations of the 5D complexities (data structure, domain, cardinality, causality, and ethics) with the phases of the data life cycle. Data agents perform tasks driven by specific <em>goals</em>. The data scientist is an abstract entity that comes from the logical organization of data agents with their actions. Data scientists face challenges that are defined according to the <em>missions</em>. We define specific discipline-induced data science, which in turn allows for the definition of <em>pan</em>-data science, a natural ecosystem that integrates specific disciplines with the essential data science. We semantically split the essential data science into computational, and foundational. By formalizing this ecosystemic view, we contribute a general-purpose, fusion-oriented architecture for integrating heterogeneous knowledge, agents, and workflows-relevant to a wide range of disciplines and high-impact applications.</div></div>","PeriodicalId":50367,"journal":{"name":"Information Fusion","volume":"130 ","pages":"Article 104113"},"PeriodicalIF":15.5,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145957302","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

SHIFT: Enhancing federated learning robustness through client-side backdoor detection SHIFT：通过客户端后门检测增强联邦学习的健壮性

IF 15.5 1区计算机科学 Q1 COMPUTER SCIENCE, ARTIFICIAL INTELLIGENCE

Information Fusion

Pub Date : 2026-06-01 Epub Date: 2026-01-10 DOI: 10.1016/j.inffus.2026.104144

Kang Wang , Liangliang Wang , Zhiquan Liu , Yiyuan Luo , Kai Zhang , Weiwei Li

Federated Learning (FL) is vulnerable to backdoor attacks, where hidden triggers in model updates can induce malicious behavior on specific inputs, ultimately compromising the reliability of FL. However, existing backdoor detection methods require decryption of locally uploaded encrypted models on the server before further detection can be performed. In this paper, we propose SHIFT containing three parts: transferring the backdoor detection task to the client side to significantly reduce the computational burden on the server; employing client-side code obfuscation to prevent malicious clients from analyzing or bypassing the detection mechanism; and utilizing a dynamic risk level mapping mechanism to adaptively adjust the results of the backdoor detection output. SHIFT can directly detect unencrypted data on the client side. We evaluated the time overhead of SHIFT compared with various backdoor detection schemes based on different encryption methods. Additionally, we assessed its performance in handwritten digit recognition and image classification tasks under single-client and multi-client backdoor attacks, specifically in non-independent and identically distributed (non-IID) scenarios. Experimental results indicate that SHIFT improves backdoor detection efficiency by a factor ranging from 1.28 to 36.65 over existing schemes, while also demonstrating robust performance in detecting and defending against various backdoor attacks, particularly in large-scale, multi-client distributed federated learning systems.

联邦学习（FL）容易受到后门攻击，其中模型更新中的隐藏触发器可能会导致特定输入的恶意行为，最终损害FL的可靠性。然而，现有的后门检测方法需要在服务器上对本地上传的加密模型进行解密，然后才能执行进一步的检测。在本文中，我们提出SHIFT包含三个部分：将后门检测任务转移到客户端，以显着减少服务器的计算负担；采用客户端代码混淆，防止恶意客户端分析或绕过检测机制；并利用动态风险等级映射机制自适应调整后门检测输出结果。SHIFT可以直接检测客户端上未加密的数据。我们评估了SHIFT与基于不同加密方法的各种后门检测方案的时间开销。此外，我们评估了其在单客户端和多客户端后门攻击下手写数字识别和图像分类任务中的性能，特别是在非独立和同分布（non-IID）场景下。实验结果表明，与现有方案相比，SHIFT将后门检测效率提高了1.28到36.65倍，同时在检测和防御各种后门攻击方面也表现出了强大的性能，特别是在大规模、多客户端分布式联邦学习系统中。

{"title":"SHIFT: Enhancing federated learning robustness through client-side backdoor detection","authors":"Kang Wang , Liangliang Wang , Zhiquan Liu , Yiyuan Luo , Kai Zhang , Weiwei Li","doi":"10.1016/j.inffus.2026.104144","DOIUrl":"10.1016/j.inffus.2026.104144","url":null,"abstract":"<div><div>Federated Learning (FL) is vulnerable to backdoor attacks, where hidden triggers in model updates can induce malicious behavior on specific inputs, ultimately compromising the reliability of FL. However, existing backdoor detection methods require decryption of locally uploaded encrypted models on the server before further detection can be performed. In this paper, we propose SHIFT containing three parts: transferring the backdoor detection task to the client side to significantly reduce the computational burden on the server; employing client-side code obfuscation to prevent malicious clients from analyzing or bypassing the detection mechanism; and utilizing a dynamic risk level mapping mechanism to adaptively adjust the results of the backdoor detection output. SHIFT can directly detect unencrypted data on the client side. We evaluated the time overhead of SHIFT compared with various backdoor detection schemes based on different encryption methods. Additionally, we assessed its performance in handwritten digit recognition and image classification tasks under single-client and multi-client backdoor attacks, specifically in non-independent and identically distributed (non-IID) scenarios. Experimental results indicate that SHIFT improves backdoor detection efficiency by a factor ranging from 1.28 to 36.65 over existing schemes, while also demonstrating robust performance in detecting and defending against various backdoor attacks, particularly in large-scale, multi-client distributed federated learning systems.</div></div>","PeriodicalId":50367,"journal":{"name":"Information Fusion","volume":"130 ","pages":"Article 104144"},"PeriodicalIF":15.5,"publicationDate":"2026-06-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145957308","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