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....

2021 Vol.34 Issue.6, Published 2021-06-25 Papers and Reports    Researches and Applications        Papers and Reports 485 Name Disambiguation Based on Heterogeneous Network Representation Learning TANG Zhengzheng, HONG Xuehai, WANG Yang, LI Yuxuan During the search for the name of an author in the system, the return of all documents of the author deteriorates the user experience. Name disambiguation can improve the retrieval accuracy. Therefore, a name disambiguation method based on heterogeneous network representation learning is proposed. Firstly, a paper heterogeneous network is constructed for each ambiguous name. Then, the representation vector of each paper node in the network is obtained based on the heterogeneous network and the Word2Vec. Finally, papers are divided up and assigned to different author entities via rule matching and a clustering method based on density with noise. The proposed method generates better performance on OAG-WholsWho competition dataset, and its effectiveness is verified. 2021 Vol. 34 (6): 485-496 [Abstract] ( 503 ) [HTML 1KB] [ PDF 1147KB] ( 324 ) 497 Multi-scale Dilated Convolutional Neural Network Model Based on Attention Mechanism WANG Jingbin, LAI Xiaolian, LEI Jing, ZHANG Jingxuan The existing temporal knowledge graph representation methods cannot capture the complex relationships within quadruple well. Most of the neural network based models are unable to model time-varying knowledge and capture rich feature information. Moreover, the interaction between entities and relations in these models is poor. Therefore, a multi-scale dilated convolutional neural network model based on attention mechanism(MSDCA) is proposed. Firstly, a time-aware relation representation is obtained using long short-term memory. Secondly, a multi-scale dilated convolutional neural network is employed to improve the interactivity of the quadruple. Finally, a multi-scale attention mechanism is utilized to capture critical features to improve completion ability of MSDCA. Link prediction experiments on multiple public temporal datasets show the superiority of MSDCA. 2021 Vol. 34 (6): 497-508 [Abstract] ( 967 ) [HTML 1KB] [ PDF 956KB] ( 696 ) 509 Chinese Portrait Painting Style Transfer Algorithm SHENG Jiachuan, DONG Yufan, LI Xiaomei, LI Yuzhi Style transfer algorithms can generate target artworks quickly. However, the problems are caused by applying style transfer algorithms directly to Chinese paintings, like uneven feature distribution and inconsistent face recognition. To address these issues, a Chinese portrait painting style transfer algorithm based on convolutional neural network(CNN) is proposed. Firstly, a brushstroke control restriction is proposed to guide the texture distribution of the image for freehand brushwork and fine brushwork of Chinese portrait painting. Then, Chinese painting moving distance is proposed to measure content and style features and transfer the style of Chinese painting to portrait photos harmoniously. Finally, the restriction for improving the loss network is put forward based on the ink tone characteristics and the blank space reservation. Experiments show that the proposed algorithm is superior in Chinese painting style and the results maintain the consistency of face recognition. 2021 Vol. 34 (6): 509-521 [Abstract] ( 564 ) [HTML 1KB] [ PDF 6286KB] ( 379 ) 522 Compound-Protein Interaction Prediction Based on Graph Attention Network and Simple Recurrent Unit LI Shuhong, JIA Lin The internal covariant shift of the data and the long distance dependence of the sequence data are not taken into account in the existing deep learning based compound protein interaction prediction methods. To solve the problem, a method for compound-protein interaction prediction based on graph attention network and simple recurrent unit is proposed. The graph attention network-gated recurrent unit is introduced to learn the graph-level representation of compound molecules, the multi-layer-simple recurrent unit is employed to learn feature vector representation of amino acid subsequences, and multilayer-feed-forward network is utilized to predict compound-protein interactions. Experiments show that the evaluation indexes of the proposed method are improved on 2 public datasets, and the effectiveness of the proposed method is verified. 2021 Vol. 34 (6): 522-531 [Abstract] ( 325 ) [HTML 1KB] [ PDF 677KB] ( 382 ) Researches and Applications 532 Network Node Completion Based on Graph Convolutional Network LIU Chen, LI Ziran, ZHOU Lixin Aiming at the incomplete network data and missing nodes in graph data structure, a network node completion algorithm based on graph convolutional network is proposed. Firstly, the observed network is sampled in pairs to construct the closed subgraph and feature matrix of the target node pair. Then, the graph convolutional neural network is employed to extract the representation vectors of subgraphs and their feature matrices for two purposes. One is to infer whether there are missing nodes between target node pairs of each subgraph, and the other is whether the missing nodes between different target node pairs are the same node. Finally, experiments on real network datasets and artificially generated network datasets show that the proposed model can solve the problem of network completion well and recover the network even when half of the nodes in the network are missing. 2021 Vol. 34 (6): 532-540 [Abstract] ( 562 ) [HTML 1KB] [ PDF 660KB] ( 602 ) 541 Soft Covering Entropy and Its Applications in Multi-attribute Group Decision-Making WU Jiaming, HUANG Zhehuang, LI Jinjin, LIU Danyue, WU Zhe Information entropy and soft coverings are combined to propose soft covering information entropy. Soft covering information entropy, soft covering joint entropy and soft covering conditional entropy are defined. The relationships between these entropies and their important properties are studied. Two kinds of algorithms for multi-attribute group decision making based on soft covering conditional entropy are presented, and the consistency between the results of these two algorithms is illustrated by examples. 2021 Vol. 34 (6): 541-551 [Abstract] ( 337 ) [HTML 1KB] [ PDF 753KB] ( 460 ) 552 Automatic Generation of Lung Description in Chest X-Ray Based on Deep Learning HUANG Xin, GU Mengdan, YI Yugen, CAO Yuanlong The chest X-ray report automatic generation is a hot research topic in computer-aided diagnosis. More than 65% of diseases in chest X-rays are related to the lungs. For the generation of Chinese reports on lung descriptions, a hierarchical long short term memory model based on semantic labels is proposed. Firstly, the abnormal chest X-ray reports are analyzed, and high-frequency keywords are extracted as semantic labels. Then, the abnormal binary-classification module is introduced to correct the semantic label classification results. Finally, semantic labels and image features are fused to enhance the association mapping between them. Experimental results show that the proposed model is superior to the baseline method in both general and domain metrics, and it improves the performance of chest radiograph report generation effectively. 2021 Vol. 34 (6): 552-560 [Abstract] ( 533 ) [HTML 1KB] [ PDF 806KB] ( 337 ) 561 Collaborative Filtering Recommendation Algorithm Based on Energy Diffusion in Social Network REN Yonggong, WANG Ruixia, ZHANG Zhipeng To improve the low accuracy caused by sparse data in the recommender system, a collaborative filtering recommendation algorithm based on energy diffusion in social networks is proposed. The abundant social information in social network and the excellent performance of energy diffusion in data sparsity are combined. Firstly, the transitivity of user-item scoring matrix and trust relationship is exploited to calculate the trust intensity value between users. Then, the resource value of items is obtained by combining the social network with the user-item binary network. Finally, the collaborative filtering method is utilized to predict the score. Experiments on real datasets show that the proposed method alleviates data sparsity and solves the problem of low recommendation accuracy. 2021 Vol. 34 (6): 561-571 [Abstract] ( 344 ) [HTML 1KB] [ PDF 688KB] ( 323 ) 572 Speech Driven Talking Face Video Generation via Landmarks Representation NIAN Fudong, WANG Wentao, WANG Yan, ZHANG Jingjing, HU Guiheng, LI Teng The speaker's head motion is ignored in the existing speech driven talking face video generation methods. Aiming at this problem, a speech driven talking face video generation method based on facial landmarks representation is proposed. The speaker's head motion information and lip motion information are represented by facial contour landmarks and lip landmarks, respectively. The speech is converted to facial landmarks through a parallel multi-branch network. The final talking face video is synthesized by continuous lip landmark sequence, head landmark sequence and template image. The corresponding quantitative and qualitative experiments are conducted. Experimental results show that the talking face video with head action synthesized by the proposed method is clear and natural, and its performance is better. 2021 Vol. 34 (6): 572-580 [Abstract] ( 559 ) [HTML 1KB] [ PDF 2728KB] ( 547 )

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