模式识别与人工智能
Home      About Journal      Editorial Board      Instructions      Ethics Statement      Contact Us                   中文
Pattern Recognition and Artificial Intelligence
22 Judgement and Disposal of Academic Misconduct Article
22 Copyright Transfer Agreement
22 Proof of Confidentiality
22 Requirements for Electronic Version
More....
22 Chinese Association of Automation
22 National ResearchCenter for Intelligent Computing System
22 Institute of Intelligent Machines,Chinese Academy of Sciences
More....
 
 
2023 Vol.36 Issue.2, Published 2023-02-25
Papers and Reports    Researches and Applications    Surveys and Reviews   
   
Papers and Reports
95 Kidney Tumor Image Segmentation Method Based on Uncertainty Guidance and Scale Consistency
HOU Bingzhen, ZHANG Guimei, PENG Kun
Aiming at the problems of boundary blurring and feature loss in kidney tumor image segmentation, a kidney tumor image segmentation method based on uncertainty guidance and scale consistency is proposed on the basis of residual attention U-net model. For blurred boundaries of kidney tumor images, an uncertainty guidance module is introduced into the decoding layer to allocate weights adaptively based on uncertainty. Thus, the effect of wrong pixels is reduced and the boundary localization ability of the model is improved. For the problem of feature loss caused by down-sampling, the scale attention module and feature consistency loss are proposed. The multi-scale fusion strategy is utilized to fuse features of different scales, and the scale consistency constraint is conducted with the features at the bottom of the encoder to alleviate the problem of feature loss. Finally, experiments of kidney and kidney tumor segmentation on the public dataset KiTS19 demonstrate that the proposed segmentation method greatly improves the segmentation accuracy. In addition, the segmentation results of the proposed method hold better reliability due to the uncertainty guidance module.
2023 Vol. 36 (2): 95-107 [Abstract] ( 512 ) [HTML 1KB] [ PDF 1603KB] ( 640 )
108 Knowledge-Guided Adaptive Sequence Reinforcement Learning Model
LI Yinggang, TONG Xiangrong
The sequence recommendation can be formalized as a Markov decision process and then transformed into a deep reinforcement learning problem. Mining critical information from user sequences is a key step, such as preference drift and dependencies between sequences. In most current deep reinforcement learning recommendation systems, a fixed sequence length is taken as the input. Inspired by knowledge graphs, a knowledge-guided adaptive sequence reinforcement learning model is proposed. Firstly, using the entity relationship of the knowledge graph, a partial sequence is intercepted from the complete user feedback sequence as a drift sequence. The item set in the drift sequence represents the user's current preference, and the sequence length represents the user's preference change speed. Then, a gated recurrent unit is utilized to extract the user's preference changes and dependencies between items, while the self-attention mechanism selectively focuses on key item information. Finally, a compound reward function is designed, including discount sequence rewards and knowledge graph rewards, to alleviate the problem of sparse reward.Experiments on four real-world datasets demonstrate that the proposed model achieves superior recommendation accuracy.
2023 Vol. 36 (2): 108-119 [Abstract] ( 343 ) [HTML 1KB] [ PDF 1074KB] ( 337 )
Surveys and Reviews
120 A Survey of Salient Object Detection Based on Scene Geometric Information
WU Lanhu, LI Zhiwei, LIU Leiye, PIAO Yongri, LU Huchuan

Salient object detection is crucial in the fields of image and video compression, camouflaged object detection, medical image segmentation, etc. With the wide application of depth sensor and light field technology, scene geometric information such as depth images and light field images is applied to salient object detection to improve the performance of the model in complex scenes. Therefore, a series of salient object detection methods based on scene geometric information is proposed. The typical algorithms of salient object detection based on scene geometric information are summarized in this paper. Firstly , the basic frameworks of models and evaluation metrics are illustrated. Then, the typical algorithms of RGB-D salient object detection and light field salient object detection are summarized and analyzed from the aspects of multi-modal feature fusion, multi-modal information optimization and network model lightweight. Meanwhile, the recent progress of salient object detection methods based on scene geometry information is introduced. Finally, the problems still faced in the existing salient object detection methods based on scene geometry information are analyzed and the future research is discussed.
2023 Vol. 36 (2): 120-142 [Abstract] ( 340 ) [HTML 1KB] [ PDF 2137KB] ( 608 )
143 Control and Decision Theory in the Metaverse: A Survey
HUANG Jie
The concept of Metaverse is widely concerned by scholars from all walks of life. Firstly, the relevant concepts of Metaverse are analyzed and introduced, and it is distinguished from similar concepts such as virtual world and digital twins. The development process of Metaverse and the necessary technologies for its establishment are combed and expounded. Assumptions for the construction levels of Metaverse in the future are put forward, and various technical support for each construction level is provided. Then, from the perspective of control science and engineering subject, the development process of control and decision theory is elaborated, and the development and application of control and decision theory in the context of Metaverse, combined with cloud computing, artificial intelligence 2.0, industrial Metaverse, blockchain and other technologies are introduced. Finally, the key challenges and openness in the theory of control and decision-making as well as the construction of Metaverse are discussed and predicted, providing inspiration and reference for future research and industrial application related to the Metaverse.
2023 Vol. 36 (2): 143-159 [Abstract] ( 286 ) [HTML 1KB] [ PDF 916KB] ( 363 )
Researches and Applications
160 Concept Factorization-Based Collaborative Multi-view Clustering Algorithm in Visible and Latent Spaces
HU Suting, SHEN Zongxin, HUANG Qianqian, HUANG Yanyong
Multi-view clustering effectively improves the clustering performance by integrating the features derived from different views. The existing multi-view clustering methods more focus on different low-dimensional representations of data and their geometrical structures in latent space, while ignoring the structural relations of data in different spaces and the clustering of different spaces. To address this issue, a concept factorization-based collaborative multi-view clustering algorithm in visible and latent spaces is proposed in this paper. Firstly, common low-dimensional feature representation of different views in latent space is extracted through concept factorization. Besides, the local structure of the original data is preserved by means of graph Laplacian regularization. Then, the data clustering in visible and latent spaces are integrated into a unified framework for collaborative learning and optimizing to obtain the final clustering results. Experimental results on eight real datasets show the superiority of the proposed method.
2023 Vol. 36 (2): 160-173 [Abstract] ( 385 ) [HTML 1KB] [ PDF 799KB] ( 337 )
174 Topic-Enhanced Multi-level Graph Neural Network for Session-Based Recommendation
TANG Gu, ZHU Xiaofei
Session-based recommendation(SBR) aims to provide recommendations for anonymous users or users who are not logged in based on data in the session. The existing research models a single item in the session as the smallest unit, ignoring the item representation in different receptive fields. Moreover, the implicit topic information contained in the session sequence is not mined. To alleviate these issues, a topic-enhanced multi-level graph neural network(TEMGNN) for SBR is proposed. Firstly, a multi-level item embedding learning module is designed to broaden the receptive fields of item and obtain the representation of items at different granularities. Then, the proposed multi-level graph neural network is employed to propagate the item information with and cross granularities, capturing richer item embedding representation. Furthermore, a topic learning module is proposed to extract the topic commonalities of items in hidden space and automatically form topic representations of items by explicit vector space projection without relying on any item attribute information. Thus, the recommendation performance of the model is enhanced. Experiments on three benchmark datasets show the superiority of TEMGNN.
2023 Vol. 36 (2): 174-186 [Abstract] ( 300 ) [HTML 1KB] [ PDF 738KB] ( 432 )
模式识别与人工智能
 

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