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....
 
 
2021 Vol.34 Issue.7, Published 2021-07-25
Papers and Reports    Researches and Applications    Metaheuristic Algorithm   
   
Metaheuristic Algorithm
581 Bean Optimization Algorithm Based on Cauchy Distribution and Parent Rotation Mechanism
LIU Hang, ZHANG Xiaoming, WANG Changjian

To improve the weak spatial exploration ability and the poor diversity of individual distribution of offspring in current bean optimization algorithm, a bean optimization algorithm based on Cauchy distribution and parent rotation mechanism(BOA-CPR) is presented. Firstly, a population distribution model based on Cauchy distribution is constructed and used in the exploration phase to improve the global search ability of the algorithm. Then, a parent rotation mechanism based on the roulette wheel selection is established to enhance the diversity of offspring. Finally, an adaptive adjustment mechanism for three important parameters, including individual distance threshold, distribution variance and descendant ratio, is designed to improve the dynamic optimization ability of the algorithm for complex optimization problems. Experiments show that BOA-CPR ranks better in average fitness value and Friedman detection index.
2021 Vol. 34 (7): 581-591 [Abstract] ( 675 ) [HTML 1KB] [ PDF 728KB] ( 432 )
592 Dual-Archive Large-Scale Sparse Optimization Algorithm Based on Dynamic Adaption
GU Qinghua, WANG Chuhao, JIANG Song, CHEN Lu
The traditional large-scale optimization algorithms generate high dimensionality and sparseness problems. A dual-archive large-scale sparse optimization algorithm based on dynamic adaptation is proposed to keep the balance of dimensionality and sparseness in the algorithm and improve the diversity and convergence performance of the algorithm in solving large-scale optimization problems. Firstly, the scores strategy for generating population is changed. By adding adaptive parameter and inertia weight, the dynamics of scores is increased, the diversity of the population is improved, and it is not easy to fall into the local optimum. Secondly, the environment selection strategy of the algorithm is changed by introducing the concept of angle truncation, and the offspring is generated effectively. Meanwhile, a double-archive strategy is introduced to separate the real decision variables from the binary decision variables and thus the running time of the algorithm is reduced. The experimental results on problems of large-scale optimization, sparse optimization and practical application show that the proposed algorithm maintains the original sparsity with steadily improved diversity and convergence and strong competitiveness.
2021 Vol. 34 (7): 592-604 [Abstract] ( 418 ) [HTML 1KB] [ PDF 1008KB] ( 435 )
605 Memory Tunicate Swarm Algorithm with Information Sharing
QU Chiwen, PENG Xiaoning
Aiming at the problems of low accuracy, slow convergence speed and easily falling into local optimum of the tunicate swarm algorithm(TSA), a memory tunicate swarm algorithm with information sharing is proposed. Firstly, a dynamic self-adaptive adjustment strategy is adopted to divide the population into two sub-groups dynamically, including information sharing search and jet propulsion search, to balance the global development capability and local development capability of TSA. Then, some tunicate individuals are selected randomly to acquire information from the peers to realize the sufficient information exchange and sharing among tunicate individuals in the information sharing search mode. For another group of individuals, historical optimal locations are introduced to guide learning and thus the effectiveness of the algorithm search is enhanced. Experimental results on 20 benchmark functions show that the proposed algorithm is evidently superior in convergence rate, solution accuracy and robustness.
2021 Vol. 34 (7): 605-618 [Abstract] ( 518 ) [HTML 1KB] [ PDF 905KB] ( 265 )
619 Multiple Populations Based Estimation of Pseudo-Normal Distribution Algorithm
YANG Qiwen, YU Shiqi, ZHANG Meilin, XUE Yuncan, CHEN Junfeng

To improve the quality of the candidate solutions and prevent the premature convergence simultaneously, a multiple populations based estimation of pseudo-normal distribution algorithm(MEPDA) is presented. The population is initialized by the good point set method and it is divided into three subgroups. By replacing sample mean with the gravity center of the samples, a pseudo-normal distribution model is obtained consequently. The probabilistic model for the subgroup evolution is built up by a linear combination of the pseudo-normal distribution models of the population and the subgroup. The comparative optimization tests on 23 benchmark functions show that MEPDA produces higher convergence speed and accuracy of the solutions. To solve the parallel assembly optimization problem with multiple constraints, the process pool, employee pool, penalty function and other measures are proposed to transform the discrete combinational optimization problem with constrained procedures and operators to an unconstrained multi-population based estimation of pseudo-normal distribution optimization problem. An engineering application demonstrates that MEPDA can be applied to the discrete combination optimization problem by just replacing the infinite set of the candidate solutions with a finite one.
2021 Vol. 34 (7): 619-630 [Abstract] ( 395 ) [HTML 1KB] [ PDF 840KB] ( 368 )
Papers and Reports
631 Variational Optical Flow Computation Method Based on Motion Optimization Semantic Segmentation
GE Liyue, DENG Shixin, GONG Jie, ZHANG Congxuan, CHEN Zhen
To address the issues of edge-blurring and over-segmentation of image sequence optical flow computation under complex scenes,such as illumination change and large displacement motions, a variational optical flow computation method based on motion optimization semantic segmentation is proposed. Firstly, an energy function of variational optical flow computation is constructed via a image local region based zero-mean normalized cross correlation matching model. Then, the motion boundary information obtained from the computed optical flows is utilized to optimize the initial image semantic segmentation result, and a variational optical flow computation model based on the motion constraint semantic segmentation is constructed. Next, the optical flows of various label areas are fused to acquirethe refined flow field. Finally, experimental results on Middlebury and UCF101 databases demonstrate that the proposed method performs well in computation accuracy and robustness, especially for the edge preserving with illumination change, textureless regions and large displacement motions.
2021 Vol. 34 (7): 631-645 [Abstract] ( 414 ) [HTML 1KB] [ PDF 4912KB] ( 573 )
646 Blurry Face Image Reconstruction Based on Asymmetric Kernel Convolution Combined with Semantic Confidence Embedding
HU Zhengping, PAN Peiyun, ZHENG Saiyue, ZHAO Mengyao, BI Shuai, LIU Yang
Inspired by the principle of heterogeneous convolution, a blurry face image reconstruction algorithm based on asymmetric kernel convolution combined with semantic confidence embedding is proposed under the framework of deep learning. Aiming at the deficiency of symmetrical square convolution kernel in expressing important features during feature extraction,asymmetric kernel is employed to replace the symmetric square convolution kernel to enhance the feature expression ability of the square convolution kernel. In the reconstruction stage, the asymmetric kernel convolution is combined with the semantic confidence network to further extract the most benificial features of each type of semantic information for the results in the reconstruction. The confidence is combined to guide the network to train in a more suitable direction for reconstruction. Experimental results on CelebA and Helen datasets show that the proposed algorithm produces better reconstruction results.
2021 Vol. 34 (7): 646-654 [Abstract] ( 333 ) [HTML 1KB] [ PDF 1284KB] ( 310 )
Researches and Applications
655 Sentencing Prediction Based onMulti-view Knowledge Graph Embedding
WANG Zhizheng, WANG Lei, LI Shuaichi, SUN Yuanyuan, CHEN Yanguang, XU Ce, WANG Gang, LIN Hongfei
Sentencing prediction is a crucial component of smart judicial construction. To make sentencing results more interpretable, the sentencing prediction task is defined as a link prediction task based on a knowledge graph. In this paper, a multi-view knowledge graph embedding method is proposed to predict the sentencing of a case. Firstly, a knowledge graph ontology pattern is designed to guide the extraction of essential elements in the case description. Next, an auxiliary graph is constructed by the extracted elements and the graph embedding method is applied to learn the initial representations of elements from this auxiliary graph. Finally, the representation of elements is enhanced by fusing the structural features of the knowledge graph. Taking drug trafficking cases as the research data, the proposed method generates better performance in sentencing prediction task based on knowledge graph, and the interpretability of sentencing results is improved.
2021 Vol. 34 (7): 655-665 [Abstract] ( 554 ) [HTML 1KB] [ PDF 1123KB] ( 390 )
666 Collaborative Filtering Recommendation Algorithm Based on Weighted Tripartite Network
REN Yonggong, WANG Ningjing, ZHANG Zhipeng
The over-concentration of recommendation results of user-based collaborative filtering algorithm on popular items causes the lack of diversity, novelty and coverage. Aiming at this problem, a collaborative filtering recommendation algorithm based on weighted tripartite network is proposed. Based on sparse analysis data and little additional information, tags are introduced to reflect user interests and item attributes simultaneously. Ternary relationships among users, items and tags are utilized to construct a tripartite network.The user preference is obtained by projecting the tripartite network to the one-mode network, and a tripartite network model weighted by user preference is constructed. According to the heat spreading method, resources are redistributed on the weighted tripartite network to find more similarity relationships. The standard framework of collaborative filtering is applied for prediction and recommendation. Experiments on real datasets show that the proposed method mines long-tail items better and realizes personalized recommendations.
2021 Vol. 34 (7): 666-676 [Abstract] ( 376 ) [HTML 1KB] [ PDF 700KB] ( 339 )
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