- Yutai Duan;Jianming Wang;Haoran Ma;Yukuan Sun;
The pooling operation is used in graph classification tasks to leverage hierarchical structures preserved in data and reduce computational complexity. However, pooling shrinkage discards graph details, and existing pooling methods may lead to the loss of key classification features. In this work, we propose a residual convolutional graph neural network to tackle the problem of key classification features losing. Particularly, our contributions are threefold:(1) Different from existing methods, we propose a new strategy to calculate sorting values and verify their importance for graph classification. Our strategy does not only use features of simple nodes but also their neighbors for the accurate evaluation of its importance.(2) We design a new graph convolutional layer architecture with the residual connection. By feeding discarded features back into the network architecture, we reduce the probability of losing critical features for graph classification.(3) We propose a new method for graph-level representation. The messages for each node are aggregated separately, and then different attention levels are assigned to each node and merged into a graph-level representation to retain structural and critical information for classification. Our experimental results show that our method leads to state-of-the-art results on multiple graph classification benchmarks.
2022年04期 v.27 653-663页 [查看摘要][在线阅读][下载 655K] [下载次数：78 ] |[网刊下载次数：0 ] |[引用频次：3 ] |[阅读次数：0 ] - Yutai Duan;Jianming Wang;Haoran Ma;Yukuan Sun;
The pooling operation is used in graph classification tasks to leverage hierarchical structures preserved in data and reduce computational complexity. However, pooling shrinkage discards graph details, and existing pooling methods may lead to the loss of key classification features. In this work, we propose a residual convolutional graph neural network to tackle the problem of key classification features losing. Particularly, our contributions are threefold:(1) Different from existing methods, we propose a new strategy to calculate sorting values and verify their importance for graph classification. Our strategy does not only use features of simple nodes but also their neighbors for the accurate evaluation of its importance.(2) We design a new graph convolutional layer architecture with the residual connection. By feeding discarded features back into the network architecture, we reduce the probability of losing critical features for graph classification.(3) We propose a new method for graph-level representation. The messages for each node are aggregated separately, and then different attention levels are assigned to each node and merged into a graph-level representation to retain structural and critical information for classification. Our experimental results show that our method leads to state-of-the-art results on multiple graph classification benchmarks.
2022年04期 v.27 653-663页 [查看摘要][在线阅读][下载 655K] [下载次数：78 ] |[网刊下载次数：0 ] |[引用频次：3 ] |[阅读次数：0 ] - Cheng Peng;Chunxia Zhang;Xiaojun Xue;Jiameng Gao;Hongjian Liang;Zhengdong Niu;
Multimodal Sentiment Classification(MSC) uses multimodal data, such as images and texts, to identify the users' sentiment polarities from the information posted by users on the Internet. MSC has attracted considerable attention because of its wide applications in social computing and opinion mining. However, improper correlation strategies can cause erroneous fusion as the texts and the images that are unrelated to each other may integrate.Moreover, simply concatenating them modal by modal, even with true correlation, cannot fully capture the features within and between modals. To solve these problems, this paper proposes a Cross-Modal Complementary Network(CMCN) with hierarchical fusion for MSC. The CMCN is designed as a hierarchical structure with three key modules,namely, the feature extraction module to extract features from texts and images, the feature attention module to learn both text and image attention features generated by an image-text correlation generator, and the cross-modal hierarchical fusion module to fuse features within and between modals. Such a CMCN provides a hierarchical fusion framework that can fully integrate different modal features and helps reduce the risk of integrating unrelated modal features. Extensive experimental results on three public datasets show that the proposed approach significantly outperforms the state-of-the-art methods.
2022年04期 v.27 664-679页 [查看摘要][在线阅读][下载 774K] [下载次数：56 ] |[网刊下载次数：0 ] |[引用频次：4 ] |[阅读次数：0 ] - Cheng Peng;Chunxia Zhang;Xiaojun Xue;Jiameng Gao;Hongjian Liang;Zhengdong Niu;
Multimodal Sentiment Classification(MSC) uses multimodal data, such as images and texts, to identify the users' sentiment polarities from the information posted by users on the Internet. MSC has attracted considerable attention because of its wide applications in social computing and opinion mining. However, improper correlation strategies can cause erroneous fusion as the texts and the images that are unrelated to each other may integrate.Moreover, simply concatenating them modal by modal, even with true correlation, cannot fully capture the features within and between modals. To solve these problems, this paper proposes a Cross-Modal Complementary Network(CMCN) with hierarchical fusion for MSC. The CMCN is designed as a hierarchical structure with three key modules,namely, the feature extraction module to extract features from texts and images, the feature attention module to learn both text and image attention features generated by an image-text correlation generator, and the cross-modal hierarchical fusion module to fuse features within and between modals. Such a CMCN provides a hierarchical fusion framework that can fully integrate different modal features and helps reduce the risk of integrating unrelated modal features. Extensive experimental results on three public datasets show that the proposed approach significantly outperforms the state-of-the-art methods.
2022年04期 v.27 664-679页 [查看摘要][在线阅读][下载 774K] [下载次数：56 ] |[网刊下载次数：0 ] |[引用频次：4 ] |[阅读次数：0 ] - Yilin Kang;Renwei Ou;Yi Zhang;Hongling Li;Shasha Tian;
Government policy-group integration and policy-chain inference are significant to the execution of strategies in current Chinese society. Specifically, the coordination of hierarchical policies implemented among government departments is one of the key challenges to rural revitalization. In recent years, various well-established quantitative methods have been proposed to evaluate policy coordination, but the majority of these relied on manual analysis,which can lead to subjective results. Thus, in this paper, a novel approach called “policy knowledge graph for the coordination among the government departments”(PG-CODE) is proposed, which incorporates topic modeling into policy knowledge graphs. Similar to a knowledge graph, a policy knowledge graph uses a graph-structured data model to integrate policy discourse. With latent Dirichlet allocation embedding, a policy knowledge graph could capture the underlying topics of the policies. Furthermore, coordination strength and topic diffusion among hierarchical departments could be inferred from the PG-CODE, as it can provide a better representation of coordination within the policy space. We implemented and evaluated the PG-CODE in the field of rural innovation and entrepreneurship policy, and the results effectively demonstrate improved coordination among departments.
2022年04期 v.27 680-691页 [查看摘要][在线阅读][下载 748K] [下载次数：155 ] |[网刊下载次数：0 ] |[引用频次：6 ] |[阅读次数：0 ] - Yilin Kang;Renwei Ou;Yi Zhang;Hongling Li;Shasha Tian;
Government policy-group integration and policy-chain inference are significant to the execution of strategies in current Chinese society. Specifically, the coordination of hierarchical policies implemented among government departments is one of the key challenges to rural revitalization. In recent years, various well-established quantitative methods have been proposed to evaluate policy coordination, but the majority of these relied on manual analysis,which can lead to subjective results. Thus, in this paper, a novel approach called “policy knowledge graph for the coordination among the government departments”(PG-CODE) is proposed, which incorporates topic modeling into policy knowledge graphs. Similar to a knowledge graph, a policy knowledge graph uses a graph-structured data model to integrate policy discourse. With latent Dirichlet allocation embedding, a policy knowledge graph could capture the underlying topics of the policies. Furthermore, coordination strength and topic diffusion among hierarchical departments could be inferred from the PG-CODE, as it can provide a better representation of coordination within the policy space. We implemented and evaluated the PG-CODE in the field of rural innovation and entrepreneurship policy, and the results effectively demonstrate improved coordination among departments.
2022年04期 v.27 680-691页 [查看摘要][在线阅读][下载 748K] [下载次数：155 ] |[网刊下载次数：0 ] |[引用频次：6 ] |[阅读次数：0 ] - Babatounde Moctard Oloulade;Jianliang Gao;Jiamin Chen;Tengfei Lyu;Raeed Al-Sabri;
In academia and industries, graph neural networks(GNNs) have emerged as a powerful approach to graph data processing ranging from node classification and link prediction tasks to graph clustering tasks. GNN models are usually handcrafted. However, building handcrafted GNN models is difficult and requires expert experience because GNN model components are complex and sensitive to variations. The complexity of GNN model components has brought significant challenges to the existing efficiencies of GNNs. Hence, many studies have focused on building automated machine learning frameworks to search for the best GNN models for targeted tasks. In this work, we provide a comprehensive review of automatic GNN model building frameworks to summarize the status of the field to facilitate future progress. We categorize the components of automatic GNN model building frameworks into three dimensions according to the challenges of building them. After reviewing the representative works for each dimension, we discuss promising future research directions in this rapidly growing field.
2022年04期 v.27 692-708页 [查看摘要][在线阅读][下载 403K] [下载次数：59 ] |[网刊下载次数：0 ] |[引用频次：6 ] |[阅读次数：0 ] - Babatounde Moctard Oloulade;Jianliang Gao;Jiamin Chen;Tengfei Lyu;Raeed Al-Sabri;
In academia and industries, graph neural networks(GNNs) have emerged as a powerful approach to graph data processing ranging from node classification and link prediction tasks to graph clustering tasks. GNN models are usually handcrafted. However, building handcrafted GNN models is difficult and requires expert experience because GNN model components are complex and sensitive to variations. The complexity of GNN model components has brought significant challenges to the existing efficiencies of GNNs. Hence, many studies have focused on building automated machine learning frameworks to search for the best GNN models for targeted tasks. In this work, we provide a comprehensive review of automatic GNN model building frameworks to summarize the status of the field to facilitate future progress. We categorize the components of automatic GNN model building frameworks into three dimensions according to the challenges of building them. After reviewing the representative works for each dimension, we discuss promising future research directions in this rapidly growing field.
2022年04期 v.27 692-708页 [查看摘要][在线阅读][下载 403K] [下载次数：59 ] |[网刊下载次数：0 ] |[引用频次：6 ] |[阅读次数：0 ] - Shengyi Jiang;Sihui Fu;Nankai Lin;Yingwen Fu;
Trained on a large corpus, pretrained models(PTMs) can capture different levels of concepts in context and hence generate universal language representations, which greatly benefit downstream natural language processing(NLP) tasks. In recent years, PTMs have been widely used in most NLP applications, especially for high-resource languages, such as English and Chinese. However, scarce resources have discouraged the progress of PTMs for low-resource languages. Transformer-based PTMs for the Khmer language are presented in this work for the first time. We evaluate our models on two downstream tasks: Part-of-speech tagging and news categorization. The dataset for the latter task is self-constructed. Experiments demonstrate the effectiveness of the Khmer models. In addition, we find that the current Khmer word segmentation technology does not aid performance improvement. We aim to release our models and datasets to the community in hopes of facilitating the future development of Khmer NLP applications.
2022年04期 v.27 709-718页 [查看摘要][在线阅读][下载 424K] [下载次数：37 ] |[网刊下载次数：0 ] |[引用频次：2 ] |[阅读次数：0 ] - Shengyi Jiang;Sihui Fu;Nankai Lin;Yingwen Fu;
Trained on a large corpus, pretrained models(PTMs) can capture different levels of concepts in context and hence generate universal language representations, which greatly benefit downstream natural language processing(NLP) tasks. In recent years, PTMs have been widely used in most NLP applications, especially for high-resource languages, such as English and Chinese. However, scarce resources have discouraged the progress of PTMs for low-resource languages. Transformer-based PTMs for the Khmer language are presented in this work for the first time. We evaluate our models on two downstream tasks: Part-of-speech tagging and news categorization. The dataset for the latter task is self-constructed. Experiments demonstrate the effectiveness of the Khmer models. In addition, we find that the current Khmer word segmentation technology does not aid performance improvement. We aim to release our models and datasets to the community in hopes of facilitating the future development of Khmer NLP applications.
2022年04期 v.27 709-718页 [查看摘要][在线阅读][下载 424K] [下载次数：37 ] |[网刊下载次数：0 ] |[引用频次：2 ] |[阅读次数：0 ] - Jianliang Gao;Xiangyue Liu;Yibo Chen;Fan Xiong;
Knowledge graphs(KGs) provide a wealth of prior knowledge for the research on social networks. Crosslingual entity alignment aims at integrating complementary KGs from different languages and thus benefits various knowledge-driven social network studies. Recent entity alignment methods often take an embedding-based approach to model the entity and relation embedding of KGs. However, these studies mostly focus on the information of the entity itself and its structural features but ignore the influence of multiple types of data in KGs. In this paper, we propose a new embedding-based framework named multiview highway graph convolutional network(MHGCN),which considers the entity alignment from the views of entity semantic, relation semantic, and entity attribute. To learn the structural features of an entity, the MHGCN employs a highway graph convolutional network(GCN) for entity embedding in each view. In addition, the MHGCN weights and fuses the multiple views according to the importance of the embedding from each view to obtain a better entity embedding. The alignment entities are identified based on the similarity of entity embeddings. The experimental results show that the MHGCN consistently outperforms the state-of-the-art alignment methods. The research also will benefit knowledge fusion through cross-lingual KG entity alignment.
2022年04期 v.27 719-728页 [查看摘要][在线阅读][下载 1597K] [下载次数：45 ] |[网刊下载次数：0 ] |[引用频次：9 ] |[阅读次数：0 ] - Jianliang Gao;Xiangyue Liu;Yibo Chen;Fan Xiong;
Knowledge graphs(KGs) provide a wealth of prior knowledge for the research on social networks. Crosslingual entity alignment aims at integrating complementary KGs from different languages and thus benefits various knowledge-driven social network studies. Recent entity alignment methods often take an embedding-based approach to model the entity and relation embedding of KGs. However, these studies mostly focus on the information of the entity itself and its structural features but ignore the influence of multiple types of data in KGs. In this paper, we propose a new embedding-based framework named multiview highway graph convolutional network(MHGCN),which considers the entity alignment from the views of entity semantic, relation semantic, and entity attribute. To learn the structural features of an entity, the MHGCN employs a highway graph convolutional network(GCN) for entity embedding in each view. In addition, the MHGCN weights and fuses the multiple views according to the importance of the embedding from each view to obtain a better entity embedding. The alignment entities are identified based on the similarity of entity embeddings. The experimental results show that the MHGCN consistently outperforms the state-of-the-art alignment methods. The research also will benefit knowledge fusion through cross-lingual KG entity alignment.
2022年04期 v.27 719-728页 [查看摘要][在线阅读][下载 1597K] [下载次数：45 ] |[网刊下载次数：0 ] |[引用频次：9 ] |[阅读次数：0 ]
- Zhen Wang;Guofa Zhang;Jing Ye;Jianhui Jiang;Fengyong Li;Yong Wang;
In recent years, Approximate Computing Circuits(ACCs) have been widely used in applications with intrinsic tolerance to errors. With the increased availability of approximate computing circuit approaches, reliability analysis methods for assessing their fault vulnerability have become highly necessary. In this study, two accurate reliability evaluation methods for approximate computing circuits are proposed. The reliability of approximate computing circuits is calculated on the basis of the iterative Probabilistic Transfer Matrix(PTM) model. During the calculation, the correlation coefficients are derived and combined to deal with the correlation problem caused by fanout reconvergence. The accuracy and scalability of the two methods are verified using three sets of approximate computing circuit instances and more circuits in Evo Approx8 b, which is an approximate computing circuit open source library. Experimental results show that relative to the Monte Carlo simulation, the two methods achieve average error rates of 0.46% and 1.29% and time overheads of 0.002% and 0.1%. Different from the existing approaches to reliability estimation for approximate computing circuits based on the original PTM model, the proposed methods reduce the space overheads by nearly 50% and achieve time overheads of 1.78% and 2.19%.
2022年04期 v.27 729-740页 [查看摘要][在线阅读][下载 387K] [下载次数：26 ] |[网刊下载次数：0 ] |[引用频次：3 ] |[阅读次数：0 ] - Zhen Wang;Guofa Zhang;Jing Ye;Jianhui Jiang;Fengyong Li;Yong Wang;
In recent years, Approximate Computing Circuits(ACCs) have been widely used in applications with intrinsic tolerance to errors. With the increased availability of approximate computing circuit approaches, reliability analysis methods for assessing their fault vulnerability have become highly necessary. In this study, two accurate reliability evaluation methods for approximate computing circuits are proposed. The reliability of approximate computing circuits is calculated on the basis of the iterative Probabilistic Transfer Matrix(PTM) model. During the calculation, the correlation coefficients are derived and combined to deal with the correlation problem caused by fanout reconvergence. The accuracy and scalability of the two methods are verified using three sets of approximate computing circuit instances and more circuits in Evo Approx8 b, which is an approximate computing circuit open source library. Experimental results show that relative to the Monte Carlo simulation, the two methods achieve average error rates of 0.46% and 1.29% and time overheads of 0.002% and 0.1%. Different from the existing approaches to reliability estimation for approximate computing circuits based on the original PTM model, the proposed methods reduce the space overheads by nearly 50% and achieve time overheads of 1.78% and 2.19%.
2022年04期 v.27 729-740页 [查看摘要][在线阅读][下载 387K] [下载次数：26 ] |[网刊下载次数：0 ] |[引用频次：3 ] |[阅读次数：0 ] - Abolfazl Amraeinia;Yuhua Zuo;Jun Zheng;Zhi Liu;Guangze Zhang;Liping Luo;Buwen Cheng;Xiaoping Zou;Chunbo Li;
Perovskite Solar Cells (PSCs) have attracted considerable attention because of their unique features and high efficiency.However,the stability of perovskite solar cells remains to be improved.In this study,we modified the TiO_(2 )Electron Transport Layer (ETL) interface with PbCl_2.The efficiency of the perovskite solar cells with carbon electrodes increased from 11.28% to 13.34%,and their stability obviously improved.The addition of PbCl_(2 )had no effect on the morphology,crystal structure,and absorption property of the perovskite absorber layer.However,it affected the band energy level alignment of the solar cells and accelerated the electron extraction and transfer at the interface between the perovskite layer and the ETL,thus enhancing the overall photovoltaic performance.The interfacial modification of ETL with PbCl_(2 )is a promising way for the potential commercialization of low-cost carbon electrode-based perovskite solar cells.
2022年04期 v.27 741-750页 [查看摘要][在线阅读][下载 2168K] [下载次数：19 ] |[网刊下载次数：0 ] |[引用频次：3 ] |[阅读次数：0 ] - Abolfazl Amraeinia;Yuhua Zuo;Jun Zheng;Zhi Liu;Guangze Zhang;Liping Luo;Buwen Cheng;Xiaoping Zou;Chunbo Li;
Perovskite Solar Cells (PSCs) have attracted considerable attention because of their unique features and high efficiency.However,the stability of perovskite solar cells remains to be improved.In this study,we modified the TiO_(2 )Electron Transport Layer (ETL) interface with PbCl_2.The efficiency of the perovskite solar cells with carbon electrodes increased from 11.28% to 13.34%,and their stability obviously improved.The addition of PbCl_(2 )had no effect on the morphology,crystal structure,and absorption property of the perovskite absorber layer.However,it affected the band energy level alignment of the solar cells and accelerated the electron extraction and transfer at the interface between the perovskite layer and the ETL,thus enhancing the overall photovoltaic performance.The interfacial modification of ETL with PbCl_(2 )is a promising way for the potential commercialization of low-cost carbon electrode-based perovskite solar cells.
2022年04期 v.27 741-750页 [查看摘要][在线阅读][下载 2168K] [下载次数：19 ] |[网刊下载次数：0 ] |[引用频次：3 ] |[阅读次数：0 ] - Xiangmin Ji;Baonan Wang;Feng Hu;Chao Wang;Huanguo Zhang;
Universal quantum computers are far from achieving practical applications. The D-Wave quantum computer is initially designed for combinatorial optimizations. Therefore, exploring the potential applications of the D-Wave device in the field of cryptography is of great importance. First, although we optimize the general quantum Hamiltonian on the basis of the structure of the multiplication table(factor up to 1 005 973), this study attempts to explore the simplification of Hamiltonian derived from the binary structure of the integers to be factored. A simple factorization on 143 with four qubits is provided to verify the potential of further advancing the integer-factoring ability of the D-Wave device. Second, by using the quantum computing cryptography based on the D-Wave 2000 Q system, this research further constructs a simple version of quantum-classical computing architecture and a Quantum-Inspired Simulated Annealing(QISA) framework. Good functions and a high-performance platform are introduced, and additional balanced Boolean functions with high nonlinearity and optimal algebraic immunity can be found. Further comparison between QISA and Quantum Annealing(QA) on six-variable bent functions not only shows the potential speedup of QA, but also suggests the potential of architecture to be a scalable way of D-Wave annealer toward a practical cryptography design.
2022年04期 v.27 751-759页 [查看摘要][在线阅读][下载 280K] [下载次数：35 ] |[网刊下载次数：0 ] |[引用频次：0 ] |[阅读次数：0 ] - Xiangmin Ji;Baonan Wang;Feng Hu;Chao Wang;Huanguo Zhang;
Universal quantum computers are far from achieving practical applications. The D-Wave quantum computer is initially designed for combinatorial optimizations. Therefore, exploring the potential applications of the D-Wave device in the field of cryptography is of great importance. First, although we optimize the general quantum Hamiltonian on the basis of the structure of the multiplication table(factor up to 1 005 973), this study attempts to explore the simplification of Hamiltonian derived from the binary structure of the integers to be factored. A simple factorization on 143 with four qubits is provided to verify the potential of further advancing the integer-factoring ability of the D-Wave device. Second, by using the quantum computing cryptography based on the D-Wave 2000 Q system, this research further constructs a simple version of quantum-classical computing architecture and a Quantum-Inspired Simulated Annealing(QISA) framework. Good functions and a high-performance platform are introduced, and additional balanced Boolean functions with high nonlinearity and optimal algebraic immunity can be found. Further comparison between QISA and Quantum Annealing(QA) on six-variable bent functions not only shows the potential speedup of QA, but also suggests the potential of architecture to be a scalable way of D-Wave annealer toward a practical cryptography design.
2022年04期 v.27 751-759页 [查看摘要][在线阅读][下载 280K] [下载次数：35 ] |[网刊下载次数：0 ] |[引用频次：0 ] |[阅读次数：0 ] - Junyu Ren;Jinze Li;Huaxing Liu;Tuanfa Qin;
With the rapid advancement of the Internet of Things(IoT), the typical application of wireless body area networks(WBANs) based smart healthcare has drawn wide attention from all sectors of society. To alleviate the pressing challenges, such as resource limitations, low-latency service provision, mass data processing, rigid security demands, and the lack of a central entity, the advanced solutions of fog computing, software-defined networking(SDN) and blockchain are leveraged in this work. On the basis of these solutions, a task offloading strategy with a centralized low-latency, secure and reliable decision-making algorithm having powerful emergency handling capacity(LSRDM-EH) is designed to facilitate the resource-constrained edge devices for task offloading. Additionally, to well ensure the security of the entire network, a comprehensive blockchain-based two-layer and multidimensional security strategy is proposed. Furthermore, to tackle the inherent time-inefficiency problem of blockchain, we propose a blockchain sharding scheme to reduce system time latency. Extensive simulation has been conducted to validate the performance of the proposed measures, and numerical results verify the superiority of our methods with lower time-latency, higher reliability and security.
2022年04期 v.27 760-776页 [查看摘要][在线阅读][下载 2118K] [下载次数：43 ] |[网刊下载次数：0 ] |[引用频次：16 ] |[阅读次数：0 ] - Junyu Ren;Jinze Li;Huaxing Liu;Tuanfa Qin;
With the rapid advancement of the Internet of Things(IoT), the typical application of wireless body area networks(WBANs) based smart healthcare has drawn wide attention from all sectors of society. To alleviate the pressing challenges, such as resource limitations, low-latency service provision, mass data processing, rigid security demands, and the lack of a central entity, the advanced solutions of fog computing, software-defined networking(SDN) and blockchain are leveraged in this work. On the basis of these solutions, a task offloading strategy with a centralized low-latency, secure and reliable decision-making algorithm having powerful emergency handling capacity(LSRDM-EH) is designed to facilitate the resource-constrained edge devices for task offloading. Additionally, to well ensure the security of the entire network, a comprehensive blockchain-based two-layer and multidimensional security strategy is proposed. Furthermore, to tackle the inherent time-inefficiency problem of blockchain, we propose a blockchain sharding scheme to reduce system time latency. Extensive simulation has been conducted to validate the performance of the proposed measures, and numerical results verify the superiority of our methods with lower time-latency, higher reliability and security.
2022年04期 v.27 760-776页 [查看摘要][在线阅读][下载 2118K] [下载次数：43 ] |[网刊下载次数：0 ] |[引用频次：16 ] |[阅读次数：0 ] 下载本期数据