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....

2022 Vol.35 Issue.8, Published 2022-08-25 Papers and Reports    Researches and Applications    Surveys and Reviews        Papers and Reports 671 Two Kinds of Variable Precision Models Based on Skill for Constructing Knowledge Structures and Skill Subset Reduction YANG Taoli, LI Jinjin, LI Zhaowen, JIN Ming, ZHOU Yinfeng, Lin Yidong It is unreasonable to evaluate the knowledge mastery of different individuals by a same knowledge structure in the existing models of constructing knowledge structures. In this paper, two problems are mainly discussed. The conditions of knowledge structure delineated via disjunctive model, conjunctive model and competency model by a skill map are too loose or harsh, and the same knowledge state may be delineated by different skill subsets on the same model. Firstly, the concepts of skill inclusion degree and competency inclusion degree are introduced, and the variable precision α-model and variable precision α-competency model for constructing knowledge structures are established. Secondly, the conditions that a skill map should satisfy while delineating well-graded knowledge structures are discussed, and the result that the knowledge structures delineated by well-skill maps are well-graded knowledge structures is obtained. Then, in view of equivalent skill subsets existing in skill maps and skill functions, skill subset reduction keeping knowledge structure unchanged and learning path selection are studied, respectively. Algorithms for acquiring the family of minimal skill subsets and knowledge structures are given. Finally, experiments on 6 datasets show the feasibility and effectiveness of the proposed algorithms. 2022 Vol. 35 (8): 671-687 [Abstract] ( 554 ) [HTML 1KB] [ PDF 951KB] ( 343 ) 688 A Multi-strategy Artificial Bee Colony Algorithm Based on Fitness Grouping ZHOU Xinyu, HU Jiancheng, WU Yanlin, ZHONG Maosheng, WANG Mingwen The multi-strategy mechanism is an effective way to improve the performance of artificial bee colony algorithm(ABC). However, characteristics of different individuals in the population are not considered in the existing methods, and the strategies are typically assigned to individuals without distinction. Consequently, the effectiveness of the multi-strategy mechanism is limited. Therefore, a multi-strategy ABC algorithm based on fitness grouping is proposed in this paper with consideration of both excellent individuals and poor individuals. Firstly, the population is divided into three groups according to fitness value of the individuals. Thus, the individuals of each group hold their own characteristics and preferences for exploration or exploitation. Then, solution search equations with distinct search capabilities are designed for three groups respectively to achieve division and cooperation among the groups and balance exploration and exploitation of the whole population. Finally, a solution search equation integrating the global best individual and some elite individuals is specially designed to further maintain the original role of the onlooker bee phase. In this scenario, the superior individuals can guide the search procedure. Experimental results on CEC2013 and CEC2015 datasets indicate the strong competitiveness of the proposed algorithm. 2022 Vol. 35 (8): 688-700 [Abstract] ( 611 ) [HTML 1KB] [ PDF 905KB] ( 422 ) Surveys and Reviews 701 Overview and Development of True 3D Display Technology: Principles and Perspectives ZHANG Mei, WANG Fei-Yue, GUO Zhen, TANG Lele, WANG Xiao The true three-dimensional(3D) display technology aims to reproduce real 3D scenes. It becomes more urgently needed than traditional 2D display systems due to the advantages of natural 3D visual perception and intuitive user experience. Since the true 3D display represents the future development trend of display area, it is the crucial technology to promote the development of metaverse area. The window of the information link between the virtual and the real world can be directly opened through the true 3D display technologies. And the immersive human-computer interaction 3D visual perception can be felt more strongly than other display technologies. In this paper, the principles and representative prototypes of various true three-dimensional display technologies are summarized. Then, the advantages and disadvantages of various main true three-dimensional display technologies are analyzed in detail. The potential development trends of the 3D display technology are proposed. Finally, the prospect forecast for true 3D display technology is provided. With the development of computer technology, optoelectronic technology, 5G communication and other technologies, the true 3D display system with high-quality display, powerful computing capabilities and intelligent perception and interactive functions will be highly developed in the future. Various applications will gradually increase in the areas of military affairs, medicine, teaching, etc. 2022 Vol. 35 (8): 701-717 [Abstract] ( 1019 ) [HTML 1KB] [ PDF 6838KB] ( 449 ) 718 A Survey of Problem Setting-Driven Deep Reinforcement Learning ZHANG Zhengfeng, ZHAO Binqi, SHAN Hongming, ZHANG Junping Combined with deep models, deep reinforcement learning(RL) is widely applied in various fields such as intelligent control and game competition. However, the existing RL surveys mainly focus on some core difficulty and neglect the analysis of problem itself from an overall perspective. The practical application in real-world scenarios is confronted with many technical challenges , and the technical approaches for a particular problem are not as good as expected for specific scenarios. Therefore, problem setting is defined in this paper from six major aspects, including agent, task distribution, Markov decision process, policy class, learning objective and interaction mode. A problem setting-driven perspective is utilized to analyze overall research status, elementary and extended RL setting. Then, development direction, key technologies and main motivation of the current deep RL are further discussed. Moreover, expert interaction is taken as an example to further analyze the development trends of the field in general from the problem setting-driven perspective. Finally, hot topics and future directions for the field are proposed. 2022 Vol. 35 (8): 718-742 [Abstract] ( 631 ) [HTML 1KB] [ PDF 1510KB] ( 788 ) Researches and Applications 743 Few-Shot Node Classification Method of Graph Adaptive Prototypical Networks GUO Ruize, WEI Wei, CUI Junbiao, FENG Kai Few-shot node classification aims to make machines recognize and classify quickly from a small number of nodes. Existing few-shot node classification models are easily affected by the inaccurate node features extracted by encoders and the intra-class outliers of query set instances in sub-tasks. Therefore, a graph adaptive prototypical networks(GAPN) model is proposed. Firstly, the nodes are embedded into the metric space by the graph encoder. Then, prototypes are computed by fusing the global importance and the local importance as weight of support set instances, and thus more robust prototypes can be learned adaptively for query set instances. Finally, the distance between the class prototypes of the adaptive task and the query set instance is calculated to generate the classification probability. By minimizing the positive marginal feedback loss between the classification probability and the true label, network parameters are updated backward and more discriminative node features can be learned. Experimental results on common graph datasets show that GAPN model yields better node classification performance. 2022 Vol. 35 (8): 743-753 [Abstract] ( 577 ) [HTML 1KB] [ PDF 876KB] ( 717 ) 754 Deep Graph Convolutional Network with Dual-Branch and Multi-interaction LOU Jiaqi, YE Hailiang, YANG Bing, LI Ming, CAO Feilong Graph neural networks show excellent performance in node classification tasks. However, how to fully obtain high-order semantic features of graph data and prevent over-smoothing is one of the key issues affecting the accuracy of node classification. Therefore, deep graph convolutional network with dual-branch and multi-interaction is constructed to enhance the ability to acquire high-order semantic features of nodes. Firstly, the graph structure is reconstructed according to the feature information of the nodes. Then, a dual-branch network architecture is established by both the original and the constructed graph structures to fully extract different high-order semantic features. A channel information interaction mechanism is designed to increase the diversity of node features by learning the information interaction of different branches. Finally, experiments on multiple benchmark datasets demonstrate that the proposed method improves the accuracies of the semi-supervised node classification tasks and alleviates the over-smoothing phenomenon effectively. 2022 Vol. 35 (8): 754-763 [Abstract] ( 661 ) [HTML 1KB] [ PDF 1151KB] ( 379 ) 764 2022 Vol. 35 (8): 764-766 [Abstract] ( 289 ) [HTML 1KB] [ PDF 251KB] ( 333 )

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