Home      About Journal      Editorial Board      Instructions      Ethics Statement      Contact Us                   中文

Judgement and Disposal of Academic Misconduct Article Copyright Transfer Agreement Proof of Confidentiality Requirements for Electronic Version More.... Chinese Association of Automation National ResearchCenter for Intelligent Computing System Institute of Intelligent Machines,Chinese Academy of Sciences More....

2024 Vol.37 Issue.6, Published 2024-06-25 Papers and Reports    Multi-prefence and Multi-modality Based Recommender Method        Multi-prefence and Multi-modality Based Recommender Method 479 Multimodal Recommendation Algorithm Based on Contrastive Learning and Semantic Enhancement ZHANG Kaihan, FENG Chenjiao, YAO Kaixuan, SONG Peng, LIANG Jiye The multimodal data of items is typically introduced into recommendation algorithms as additional auxiliary information to enrich the representation features of users and items. How to effectively integrate the interaction information with multimodal information of users and items is a key issue to the research. Existing methods are still insufficient in feature fusion and semantic association modeling. Therefore, a multimodal recommendation algorithm based on contrastive learning and semantic enhancement is proposed from the perspective of feature fusion. Firstly, the graph neural network and attention mechanism are adopted to fully integrate collaborative features and multimodal features. Next, the semantic association structures within each modality are learned under the guidance of the interaction structure in collaborative information. Meanwhile, the contrastive learning paradigm is employed to capture cross-modal representation dependencies. A reliability factor is introduced into the contrastive loss to adaptively adjust the constraint strength of the multimodal features, consequently suppressing the influence of data noise. Finally, the aforementioned tasks are jointly optimized to generate recommendation results. Experimental results on four real datasets show that the proposed algorithm yields excellent performance. 2024 Vol. 37 (6): 479-490 [Abstract] ( 357 ) [HTML 1KB] [ PDF 1043KB] ( 448 ) 491 Neighborhood Extension Mechanism Enhanced Graph Parallel Focused Attention Networks for Social Recommender System LI Weiyue, ZHU Zhiguo, DONG Hao, GAO Ming, ZHANG Jun, LIU Zilong Social recommender systems are designed to predict the ratings of users for unexplored items based on their historical ratings and social connections. Most existing social recommender systems are based on graph neural networks. However, the inefficiency of attention mechanisms and the over-smoothing problem limit the precision and interpretability of rating predictions. Therefore, a neighborhood extension mechanism enhanced graph parallel focused attention network is proposed to address these issues. The overall preferences of users are decomposed into nuanced facets and a focused attention mechanism is introduced as message passing algorithm to pinpoint the item most aligned with the preferences of users based on their interaction history. Meanwhile, the mechanism identifies trustworthy friends from the social network based on diverse preferences. Furthermore, a neighborhood extension mechanism is proposed, which establishes quick link to facilitate the direct message passing between central and higher-order nodes, effectively enhancing the ability of graph focused attention network to capture the social information in higher-order ego network. Experimental results on three public benchmark datasets demonstrate the superiority of the proposed system in accurate rating prediction. Moreover, a series of visual case studies illustrate the interpretability of the system. The code for this paper can be found at: https://github.com/usernameAI/NEGA. 2024 Vol. 37 (6): 491-512 [Abstract] ( 130 ) [HTML 1KB] [ PDF 1279KB] ( 385 ) 513 Global Consistency Augmented Multi-preference Session-Based Recommendation Model WU Jiangming, ZHANG Xiaokun, XU Bo, YANG Liang, LIN Hongfei Session-based recommendation aims to predict the next item which a user is likely to interact with based on an anonymous session. However, existing session-based recommendation methods based on graph neural networks underutilize the global information. To address this issue, a global consistency augmented multi-preference session-based recommendation model(GCAM) is proposed. Firstly, a consistent global graph is constructed through the shortest path routing algorithm. The consistency of global information is ensured by capturing reliable item relationships and filtering out unreliable item relationships. Secondly, a multi-preference label smoothing strategy is applied to mine collaborative information from historical sessions to soften labels, and thereby the label can fit the true user preferences. Extensive experiments on three different datasets demonstrate the superiority of GCAM. 2024 Vol. 37 (6): 513-524 [Abstract] ( 136 ) [HTML 1KB] [ PDF 900KB] ( 348 ) 525 Sequential Recommendation Model Based on Smoothing Graph Masked Encoder LIU Yang, XIA Hongbin, LIU Yuan Aiming at the performance degradation problem of existing sequential recommendation models caused by label sparsity and user data noise, a sequential recommendation model based on smoothing graph masked encoder(SGMERec) is proposed. Firstly, a data smoothing encoder is designed to process the data, improve data quality and reduce the negative impact of extreme values and data noise. Secondly, a graph masked encoder is designed to adaptively extract transformation information from global items and a relational graph is constructed to help the model complete the missing label data, thereby enhancing the ability to deal with issues of label scarcity. Finally, batch normalization is employed to normalize the input distribution of each neural network layer. Thus, the stability of input distribution for each layer is guaranteed and the proportion of scarce labels in user sequences is reduced. Experimental results on three real datasets indicate the performance improvement of SGMERec. 2024 Vol. 37 (6): 525-537 [Abstract] ( 116 ) [HTML 1KB] [ PDF 790KB] ( 329 ) Papers and Reports 538 Optimal Scale Combinations and Attribute Reduction for Consistent Generalized Multi-scale Ordered Fuzzy Decision Systems ZHU Kang, WU Weizhi, LIU Mengxin Aiming at knowledge acquisition in generalized multi-scale ordered fuzzy decision systems, dominance relations in generalized multi-scale ordered fuzzy decision systems are firstly defined, information granules with different scale combinations in these systems are then constructed. Lower and upper approximations of sets with respect to dominance relations determined by an attribute set under different scale combinations are also defined. Five concepts of optimal scale combinations in consistent generalized multi-scale ordered fuzzy decision systems are defined. The numerical characteristics of these optimal scale combinations are described by belief and plausibility functions in the evidence theory. It is proved that belief optimal scale combinations are equivalent to lower approximate optimal scale combinations, and plausibility optimal scale combinations are equivalent to upper approximate optimal scale combinations. An attribute reduction approach based on a belief optimal scale combination is explored, and optimal scale combinations and attribute reduction search algorithms are formulated. Finally, experiments on UCI datasets verify the feasibility and validity of the proposed method and algorithms. 2024 Vol. 37 (6): 538-556 [Abstract] ( 134 ) [HTML 1KB] [ PDF 799KB] ( 323 ) 557 Multi-granularity Dynamic Scene Image Deblurring Network Based on Deep Fusion of Frequency Domain and Spatial Domain Features CHEN Zihan, ZHANG Hongyun, MIAO Duoqian, CAI Kecan Dynamic scene image deblurring is highly ill-posed, and the relative motion between the camera and the photographed target makes the blur non-uniform. Most existing deep learning methods focus on the spatial domain processing and neglect the potential contribution of the frequency domain to structural and detail recovery, leading to poor deblurring results. To solve the problems, the role of frequency domain information in image deblurring is rethought, and multi-granularity dynamic scene image deblurring network based on deep fusion of frequency domain and spatial domain features is proposed. Firstly, a frequency domain gated frequency-spatial feature deep fusion module is proposed to fully explore the correlation between spatial domain and frequency domain information. The redundancy of the fused features is reduced and the complementarity between the two domains is enhanced. Secondly, based on the proposed module, a multi-granularity network is constructed, and it fully utilizes different granularity information in the spatial domain and frequency domain for coarse-to-fine image deblurring. Finally, to solve the problem of frequency domain feature map resolution mismatch caused by different input feature map sizes during training and testing, a frequency domain resolution adaptive testing strategy is adopted to maintain the consistency of frequency changes. Experiments conducted on synthetic datasets, GoPro and HIDE, and a real dataset, RealBlur, demonstrate the proposed algorithm outperforms existing advanced algorithms in reconstructing clear images with competitive parameters and efficiency. 2024 Vol. 37 (6): 557-569 [Abstract] ( 190 ) [HTML 1KB] [ PDF 4922KB] ( 410 )

Supervised by China Association for Science and Technology Sponsored by Chinese Association of Automation NationalResearchCenter for Intelligent Computing System Institute of Intelligent Machines, Chinese Academy of Sciences Published by Science Press

Copyright © 2010 Editorial Office of Pattern Recognition and Artificial Intelligence Address: No.350 Shushanhu Road, Hefei, Anhui Province, P.R. China Tel: 0551-65591176 Fax:0551-65591176 Email: bjb@iim.ac.cn Supported by Beijing Magtech 　Email:support@magtech.com.cn