Tsinghua Science and Technology

SPECIAL SECTION ON PARALLEL AND DISTRIBUTED COMPUTING,APPLICATIONS AND TECHNOLOGIES

  • Approximation Algorithm for the Balanced 2-Correlation Clustering Problem

    Sai Ji;Dachuan Xu;Donglei Du;Ling Gai;Zhongrui Zhao;

    The Correlation Clustering Problem(CorCP) is a significant clustering problem based on the similarity of data.It has significant applications in different fields,such as machine learning,biology,and data mining,and many different problems in other areas.In this paper,the Balanced 2-CorCP(B2-CorCP) is introduced and examined,and a new interesting variant of the CorCP is described.The goal of this clustering problem is to partition the vertex set into two clusters with equal size,such that the number of disagreements is minimized.We first present a polynomial time algorithm for the B2-CorCP on M-positive edge dominant graphs(M≥ 3).Then,we provide a series of numerical experiments,and the results show the effectiveness of our algorithm.

    2022年05期 v.27 777-784页 [查看摘要][在线阅读][下载 443K]
    [下载次数:18 ] |[网刊下载次数:0 ] |[引用频次:0 ] |[阅读次数:0 ]
  • Algorithms for the Prize-Collecting k-Steiner Tree Problem

    Lu Han;Changjun Wang;Dachuan Xu;Dongmei Zhang;

    In this paper,we study the prize-collecting k-Steiner tree(PCkST) problem.We are given a graph G=(V,E) and an integer k.The graph is connected and undirected.A vertex r ∈ V called root and a subset R?V called terminals are also given.A feasible solution for the PCkST is a tree F rooted at r and connecting at least k vertices in R.Excluding a vertex from the tree incurs a penalty cost,and including an edge in the tree incurs an edge cost.We wish to find a feasible solution with minimum total cost.The total cost of a tree F is the sum of the edge costs of the edges in F and the penalty costs of the vertices not in F.We present a simple approximation algorithm with the ratio of 5.9672 for the PCkST.This algorithm uses the approximation algorithms for the prize-collecting Steiner tree(PCST) problem and the k-Steiner tree(kST) problem as subroutines.Then we propose a primal-dual based approximation algorithm and improve the approximation ratio to 5.

    2022年05期 v.27 785-792页 [查看摘要][在线阅读][下载 308K]
    [下载次数:43 ] |[网刊下载次数:0 ] |[引用频次:1 ] |[阅读次数:0 ]
  • New Benchmark for Household Garbage Image Recognition

    Zhize Wu;Huanyi Li;Xiaofeng Wang;Zijun Wu;Le Zou;Lixiang Xu;Ming Tan;

    Household garbage images are usually faced with complex backgrounds,variable illuminations,diverse angles,and changeable shapes,which bring a great difficulty in garbage image classification.Due to the ability to discover problem-specific features,deep learning and especially convolutional neural networks(CNNs) have been successfully and widely used for image representation learning.However,available and stable household garbage datasets are insufficient,which seriously limits the development of research and application.Besides,the state-of-the-art in the field of garbage image classification is not entirely clear.To solve this problem,in this study,we built a new open benchmark dataset for household garbage image classification by simulating different lightings,backgrounds,angles,and shapes.This dataset is named 30 classes of household garbage images(HGI-30),which contains 18 000 images of 30 household garbage classes.The publicly available HGI-30 dataset allows researchers to develop accurate and robust methods for household garbage recognition.We also conducted experiments and performance analyses of the state-of-the-art deep CNN methods on HGI-30,which serves as baseline results on this benchmark.

    2022年05期 v.27 793-803页 [查看摘要][在线阅读][下载 1727K]
    [下载次数:63 ] |[网刊下载次数:0 ] |[引用频次:5 ] |[阅读次数:0 ]
  • A Pixel–Channel Hybrid Attention Model for Image Processing

    Qiang Hua;Liyou Chen;Pan Li;Shipeng Zhao;Yan Li;

    In the field of image processing,better results can often be achieved through the deepening of neural network layers involving considerably more parameters.In image classification,improving classification accuracy without introducing too many parameters remains a challenge.As for image conversion,the use of the conversion model of the generative adversarial network often produces semantic artifacts,resulting in images with lower quality.Thus,to address the above problems,a new type of attention module is proposed in this paper for the first time.This proposed approach uses the pixel-channel hybrid attention(PCHA) mechanism,which combines the attention information of the pixel and channel domains.The comparative results of using different attention modules on multiple-image data verify the superiority of the PCHA module in performing classification tasks.For image conversion,we propose a skip structure(S-PCHA model) in the up-and down-sampling processes based on the PCHA model.The proposed model can help the algorithm identify the most distinctive semantic object in a given image,as this structure effectively realizes the intercommunication of encoder and decoder information.Furthermore,the results showed that the attention model could establish a more realistic mapping from the source domain to the target domain in the image conversion algorithm,thus improving the quality of the image generated by the conversion model.

    2022年05期 v.27 804-816页 [查看摘要][在线阅读][下载 1258K]
    [下载次数:68 ] |[网刊下载次数:0 ] |[引用频次:1 ] |[阅读次数:0 ]
  • Distributed Consensus for Blockchains in Internet-of-Things Networks

    Li Yang;Yifei Zou;Minghui Xu;Yicheng Xu;Dongxiao Yu;Xiuzhen Cheng;

    In recent years,due to the wide implementation of mobile agents,the Internet-of-Things(IoT) networks have been applied in several real-life scenarios,servicing applications in the areas of public safety,proximity-based services,and fog computing.Meanwhile,when more complex tasks are processed in IoT networks,demands on identity authentication,certifiable traceability,and privacy protection for services in IoT networks increase.Building a blockchain system in IoT networks can greatly satisfy such demands.However,the blockchain building in IoT brings about new challenges compared with that in the traditional full-blown Internet with reliable transmissions,especially in terms of achieving consensus on each block in complex wireless environments,which directly motivates our work.In this study,we fully considered the challenges of achieving a consensus in a blockchain system in IoT networks,including the negative impacts caused by contention and interference in wireless channel,and the lack of reliable transmissions and prior network organizations.By proposing a distributed consensus algorithm for blockchains on multi-hop IoT networks,we showed that it is possible to directly reach a consensus for blockchains in IoT networks,without relying on any additional network layers or protocols to provide reliable and ordered communications.In our theoretical analysis,we showed that our consensus algorithm is asymptotically optimal on time complexity and is energy saving.The extensive simulation results also validate our conclusions in the theoretical analysis.

    2022年05期 v.27 817-831页 [查看摘要][在线阅读][下载 1130K]
    [下载次数:23 ] |[网刊下载次数:0 ] |[引用频次:3 ] |[阅读次数:0 ]
  • Efficient Algorithms for Maximizing Group Influence in Social Networks

    Peihuang Huang;Longkun Guo;Yuting Zhong;

    In social network applications,individual opinion is often influenced by groups,and most decisions usually reflect the majority's opinions.This imposes the group influence maximization(GIM) problem that selects k initial nodes,where each node belongs to multiple groups for a given social network and each group has a weight,to maximize the weight of the eventually activated groups.The GIM problem is apparently NP-hard,given the NP-hardness of the influence maximization(IM) problem that does not consider groups.Focusing on activating groups rather than individuals,this paper proposes the complementary maximum coverage(CMC) algorithm,which greedily and iteratively removes the node with the approximate least group influence until at most k nodes remain.Although the evaluation of the current group influence against each node is only approximate,it nevertheless ensures the success of activating an approximate maximum number of groups.Moreover,we also propose the improved reverse influence sampling(IRIS) algorithm through fine-tuning of the renowned reverse influence sampling algorithm for GIM.Finally,we carry out experiments to evaluate CMC and IRIS,demonstrating that they both outperform the baseline algorithms respective of their average number of activated groups under the independent cascade(IC)model.

    2022年05期 v.27 832-842页 [查看摘要][在线阅读][下载 579K]
    [下载次数:20 ] |[网刊下载次数:0 ] |[引用频次:1 ] |[阅读次数:0 ]
  • Protein Residue Contact Prediction Based on Deep Learning and Massive Statistical Features from Multi-Sequence Alignment

    Huiling Zhang;Min Hao;Hao Wu;Hing-Fung Ting;Yihong Tang;Wenhui Xi;Yanjie Wei;

    Sequence-based protein tertiary structure prediction is of fundamental importance because the function of a protein ultimately depends on its 3 D structure.An accurate residue-residue contact map is one of the essential elements for current ab initio prediction protocols of 3 D structure prediction.Recently,with the combination of deep learning and direct coupling techniques,the performance of residue contact prediction has achieved significant progress.However,a considerable number of current Deep-Learning(DL)-based prediction methods are usually time-consuming,mainly because they rely on different categories of data types and third-party programs.In this research,we transformed the complex biological problem into a pure computational problem through statistics and artificial intelligence.We have accordingly proposed a feature extraction method to obtain various categories of statistical information from only the multi-sequence alignment,followed by training a DL model for residue-residue contact prediction based on the massive statistical information.The proposed method is robust in terms of different test sets,showed high reliability on model confidence score,could obtain high computational efficiency and achieve comparable prediction precisions with DL methods that relying on multi-source inputs.

    2022年05期 v.27 843-854页 [查看摘要][在线阅读][下载 464K]
    [下载次数:30 ] |[网刊下载次数:0 ] |[引用频次:2 ] |[阅读次数:0 ]
  • Graph Convolutional Network Combined with Semantic Feature Guidance for Deep Clustering

    Junfen Chen;Jie Han;Xiangjie Meng;Yan Li;Haifeng Li;

    The performances of semisupervised clustering for unlabeled data are often superior to those of unsupervised learning,which indicates that semantic information attached to clusters can significantly improve feature representation capability.In a graph convolutional network(GCN),each node contains information about itself and its neighbors that is beneficial to common and unique features among samples.Combining these findings,we propose a deep clustering method based on GCN and semantic feature guidance(GFDC) in which a deep convolutional network is used as a feature generator,and a GCN with a softmax layer performs clustering assignment.First,the diversity and amount of input information are enhanced to generate highly useful representations for downstream tasks.Subsequently,the topological graph is constructed to express the spatial relationship of features.For a pair of datasets,feature correspondence constraints are used to regularize clustering loss,and clustering outputs are iteratively optimized.Three external evaluation indicators,i.e.,clustering accuracy,normalized mutual information,and the adjusted Rand index,and an internal indicator,i.e., the Davidson-Bouldin index(DBI),are employed to evaluate clustering performances.Experimental results on eight public datasets show that the GFDC algorithm is significantly better than the majority of competitive clustering methods,i.e.,its clustering accuracy is20% higher than the best clustering method on the United States Postal Service dataset.The GFDC algorithm also has the highest accuracy on the smaller Amazon and Caltech datasets.Moreover,DBI indicates the dispersion of cluster distribution and compactness within the cluster.

    2022年05期 v.27 855-868页 [查看摘要][在线阅读][下载 1056K]
    [下载次数:45 ] |[网刊下载次数:0 ] |[引用频次:5 ] |[阅读次数:0 ]

REGULAR ARTICLE

  • Object Counting Using a Refinement Network

    Lehan Sun;Junjie Ma;Liping Jing;

    To address the scale variance and uneven distribution of objects in scenarios of object-counting tasks,an algorithm called Refinement Network(RefNet) is exploited.The proposed top-down scheme sequentially aggregates multiscale features,which are laterally connected with low-level information.Trained by a multiresolution density regression loss,a set of intermediate-density maps are estimated on each scale in a multiscale feature pyramid,and the detailed information of the density map is gradually added through coarse-to-fine granular refinement progress to predict the final density map.We evaluate our RefNet on three crowd-counting benchmark datasets,namely,ShanghaiTech,UCF_CC_50,and UCSD,and our method achieves competitive performances on the mean absolute error and root mean squared error compared to the state-of-the-art approaches.We further extend our RefNet to cell counting,illustrating its effectiveness on relative counting tasks.

    2022年05期 v.27 869-879页 [查看摘要][在线阅读][下载 1776K]
    [下载次数:25 ] |[网刊下载次数:0 ] |[引用频次:0 ] |[阅读次数:0 ]
  • 下载本期数据