ISSN 1000-1239 CN 11-1777/TP

    null
    null
    null

    Default Latest Most Read
    Please wait a minute...
    For Selected: Toggle Thumbnails
    Journal of Computer Research and Development    2018, 55 (9): 1827-1828.  
    Abstract479)   HTML4)    PDF (436KB)(245)       Save
    Related Articles | Metrics
    A Survey on Unsupervised Image Retrieval Using Deep Features
    Zhang Hao,Wu Jianxin
    Journal of Computer Research and Development    2018, 55 (9): 1829-1842.   DOI: 10.7544/issn1000-1239.2018.20180058
    Abstract2013)   HTML26)    PDF (2841KB)(1183)       Save
    Content-based image retrieval (CBIR) is a challenging task in computer vision. Its goal is to find images among the database images which contain the same instance as the query image. A typical image retrieval approach contains two steps: extract a proper representation vector from each raw image, and then retrieve via nearest neighbor search on those representations. The quality of the image representation vector extracted from raw image is the key factor to determine the overall performance of an image retrieval approach. Image retrieval have witnessed two developing stages, namely hand-craft feature based approaches and deep feature based approaches. Furthermore, there are two phases in each stage, i.e., one phase of using global feature and another phase of using local feature based approaches. Due to the limited representation power of hand-craft features, nowadays, the research focus of image retrieval has shifted to how to make the full utility of deep features. In this study, we give a brief review of the development progress of unsupervised image retrieval based on different ways to extract image representations. Several representative unsupervised image retrieval approaches are then introduced and compared on benchmark image retrieval datasets. At last, we discuss a few future research perspectives.
    Related Articles | Metrics
    Visual Analysis for Anomaly Detection in Time-Series: A Survey
    Han Dongming,Guo Fangzhou,Pan Jiacheng,Zheng Wenting,Chen Wei
    Journal of Computer Research and Development    2018, 55 (9): 1843-1852.   DOI: 10.7544/issn1000-1239.2018.20180126
    Abstract1671)   HTML49)    PDF (2688KB)(846)       Save
    Anomaly detection for time-series denotes the detection and analysis of abnormal and unusual patterns, trends and features. Automatic methods sometimes fail to detect anomalies that are subtle, fuzzy or uncertain, while visual analysis can overcome this challenge by integrating the capability of human users and data mining approaches through visual representations of the data and visual interface. In this paper, we identify the challenges of anomaly detection, and describe the existing works of visual analysis along two categories: types of anomalies (attributes, topologies and hybrids), and anomaly detection means (direct projection, clustering and machine learning). We highlight future research directions.
    Related Articles | Metrics
    Blockchain Data Analysis: A Review of Status, Trends and Challenges
    Chen Weili,Zheng Zibin
    Journal of Computer Research and Development    2018, 55 (9): 1853-1870.   DOI: 10.7544/issn1000-1239.2018.20180127
    Abstract4990)   HTML136)    PDF (3117KB)(2983)       Save
    Blockchain technology is a new emerging technology that has the potential to revolutionize many traditional industries. Since the creation of Bitcoin, which represents blockchain 1.0, blockchain technology has been attracting extensive attention and a great amount of user transaction data has been accumulated. Furthermore, the birth of Ethereum, which represents blockchain 2.0, further enriches data type in blockchain. While the popularity of blockchain technology bringing about a lot of technical innovation, it also leads to many new problems, such as user privacy disclosure and illegal financial activities. However, the public accessible of blockchain data provides unprecedented opportunity for researchers to understand and resolve these problems through blockchain data analysis. Thus, it is of great significance to summarize the existing research problems, the results obtained, the possible research trends, and the challenges faced in blockchain data analysis. To this end, a comprehensive review and summary of the progress of blockchain data analysis is presented. The review begins by introducing the architecture and key techniques of blockchain technology and providing the main data types in blockchain with the corresponding analysis methods. Then, the current research progress in blockchain data analysis is summarized in seven research problems, which includes entity recognition, privacy disclosure risk analysis, network portrait, network visualization, market effect analysis, transaction pattern recognition, illegal behavior detection and analysis. Finally, the directions, prospects and challenges for future research are explored based on the shortcomings of current research.
    Related Articles | Metrics
    Deep Neural Network Compression and Acceleration: A Review
    Ji Rongrong,Lin Shaohui,Chao Fei,Wu Yongjian,Huang Feiyue
    Journal of Computer Research and Development    2018, 55 (9): 1871-1888.   DOI: 10.7544/issn1000-1239.2018.20180129
    Abstract2372)   HTML36)    PDF (4080KB)(1542)       Save
    In recent years, deep neural networks (DNNs) have achieved remarkable success in many artificial intelligence (AI) applications, including computer vision, speech recognition and natural language processing. However, such DNNs have been accompanied by significant increase in computational costs and storage services, which prohibits the usages of DNNs on resource-limited environments such as mobile or embedded devices. To this end, the studies of DNN compression and acceleration have recently become more emerging. In this paper, we provide a review on the existing representative DNN compression and acceleration methods, including parameter pruning, parameter sharing, low-rank decomposition, compact filter designed, and knowledge distillation. Specifically, this paper provides an overview of DNNs, describes the details of different DNN compression and acceleration methods, and highlights the properties, advantages and drawbacks. Furthermore, we summarize the evaluation criteria and datasets widely used in DNN compression and acceleration, and also discuss the performance of the representative methods. In the end, we discuss how to choose different compression and acceleration methods to meet the needs of different tasks, and envision future directions on this topic.
    Related Articles | Metrics
    Survey of Query Processing and Mining Techniques over Large Temporal Graph Database
    Wang Yishu,Yuan Ye,Liu Meng,Wang Guoren
    Journal of Computer Research and Development    2018, 55 (9): 1889-1902.   DOI: 10.7544/issn1000-1239.2018.20180132
    Abstract1211)   HTML13)    PDF (2466KB)(606)       Save
    A temporal graph, as a graph structure with time dimension, plays a more and more important role in query processing and mining of graph data. Different with the traditional static graph, structure of the temporal graph changes with the time series, that is to say the edge of temporal graph is activated by time. And each edge of the temporal graph has the label of recording time, which makes the temporal graph contain more information than the static graph, so the existing data query processing methods cannot be used in the temporal graph. Therefore how to solve the problem of query processing and mining on the temporal graph has attracted much attention of researchers. This paper summarizes the existing query processing and mining methods on temporal graphs. Firstly, this paper gives the application background and basic definition of temporal graph, and combs the existing three typical models which are used to model temporal graph in the existing works. Secondly, this paper introduces and analyzes the existing work on temporal graph from three aspects: graph query processing method, graph mining method and temporal graph management system. Finally, the possible research directions on temporal graph are prospected to provide reference for related research.
    Related Articles | Metrics
    A Survey on Scholar Profiling Techniques in the Open Internet
    Yuan Sha, Tang Jie, Gu Xiaotao
    Journal of Computer Research and Development    2018, 55 (9): 1903-1919.   DOI: 10.7544/issn1000-1239.2018.20180139
    Abstract1612)   HTML18)    PDF (3315KB)(874)       Save
    Scholar profiling from the open Internet has become a hot research topic in recent years. Its goal is to extract the attribute information of a scholar. Scholar profiling is a fundamental issue in large-scale expert databases for finding experts, evaluating academic influence, and so on. In the open Internet, scholar profiling faces new challenges, such as large amount of data, data noise and data redundancy. The traditional user profiling methods and algorithms cannot be directly used in the user profiling system in the open Internet environment. In this paper, the existing technologies are summarized and classified to provide reference for further research. Firstly, we analyze the problem of scholar profiling, and give a general overview of the information extraction method, which is the basic theory of user profiling. Then, the three basic tasks of scholar profiling including scholar information annotation, research interest mining and academic impact prediction are introduced in detail. What’s more, the successful application system of scholar profiling called AMiner is introduced. Finally, open research issues are discussed and possible future research directions are prospected.
    Related Articles | Metrics
    Recent Advances in Datacenter Flow Scheduling
    Hu Zhiyao, Li Dongsheng, Li Ziyang
    Journal of Computer Research and Development    2018, 55 (9): 1920-1930.   DOI: 10.7544/issn1000-1239.2018.20180156
    Abstract1113)   HTML10)    PDF (2375KB)(641)       Save
    Flow scheduling techniques impose an important impact on the performance of the data center. Flow scheduling techniques aim at optimizing the user experience by controlling and scheduling the transmission link, priority and transmission rate of data flows. Flow scheduling techniques can achieve various optimization objects such as reducing the average or weighted flow completion time, decreasing the delay of long-tail flows, optimizing the transmission of flows with deadline constraints, improving the utilization of the network link. In this paper, we mainly review the recent research involving flow scheduling techniques. First, we briefly introduce data center and flow scheduling problem and challenges. These challenges mainly lie in the means to implement flow scheduling on network devices or terminal hosts, and how to design low-overhead highly-efficient scheduling algorithms. Especially, the coflow scheduling problem is proved NP-Hard to solve. Then, we review the latest progress of flow scheduling techniques from two aspects, i.e., single-flow scheduling and coflow scheduling. The divergence between single-flow scheduling techniques and coflow scheduling techniques is the flow relationship under different applications like Web search and big data analytics. In the end of the paper, we outlook the future development direction and point out some unsolved problems involving flow scheduling.
    Related Articles | Metrics
    Recent Progress in Low Differential Uniformity Functions over Finite Fields
    Qu Longjiang, Chen Xi, Niu Tailin, Li Chao
    Journal of Computer Research and Development    2018, 55 (9): 1931-1945.   DOI: 10.7544/issn1000-1239.2018.20180159
    Abstract775)   HTML2)    PDF (1305KB)(282)       Save
    To prevent differential attack on the cipher, cryptographic functions are required to have low differential uniformity. Perfect nonlinear (PN) functions, almost perfect nonlinear (APN) functions and differentially 4-uniform permutations are the most important cryptographic functions with low differential uniformity. Here we survey the recent main research results about cryptographic functions with low differential uniformity such as PN functions, APN functions and differentially 4-uniform permutations. First, we recall the connections between PN functions and the mathematical objects such as the semifield, which survey the known constructions of PN functions and the pseudo-planar functions. Second, the properties and judgement of APN functions are analyzed. We also list the known constructions of APN functions and recall the inequivalent results between them. Third, we summarize the known results on the constructions of differentially 4-uniform permutations and discuss their equivalence. Then, we recall the applications of low differential uniformity functions in the design of actual ciphers. Lastly, we propose some research problems on cryptographic functions with low differential uniformity.
    Related Articles | Metrics
    Research on Visual Question Answering Techniques
    Yu Jun, Wang Liang, Yu Zhou
    Journal of Computer Research and Development    2018, 55 (9): 1946-1958.   DOI: 10.7544/issn1000-1239.2018.20180168
    Abstract2002)   HTML35)    PDF (1926KB)(886)       Save
    With the significant advances of deep learning in computer vision and natural language processing, the existing methods are able to accurately understand the semantics of visual contents and natural languages, and carry out research on cross-media data representation and interaction. In recent years, visual question answering (VQA) has become a hot spot in cross-media expression and interaction area. The target of VQA is to learn a model to understand the visual content referred by a natural language question, and answer it automatically. This paper summarizes the research progresses in recent years on VQA from the aspects of concepts, models and datasets, and discusses the shortcomings of the current works. Finally, the possible future directions of VQA are discussed on methodology, applications and platforms.
    Related Articles | Metrics
    A Two-Stage Community Detection Algorithm Based on Label Propagation
    Zheng Wenping, Che Chenhao, Qian Yuhua, Wang Jie
    Journal of Computer Research and Development    2018, 55 (9): 1959-1971.   DOI: 10.7544/issn1000-1239.2018.20180277
    Abstract970)   HTML8)    PDF (4361KB)(540)       Save
    Due to the random process in node selection and label propagation, the stability of the results of traditional LPA is poor. A two-stage community detection algorithm is proposed based on label propagation, abbreviated as LPA-TS. In the first step of LPA-TS, the labels of nodes are updated according to their participation coefficients in non-decreasing order, and the node label is determined according to the similarity of nodes in the process of node label updating. Some clusters found by Step1 might not satisfy the weak community condition. If a cluster is not a weak community, in the beginning of Step2, we will merge it with the cluster that has most connections with it. Next, we treat each community as a node, and the number of edges between two communities as their edge weights between corresponding nodes. We compute the participation coefficients of each node of the resulted network, and use similar process to get the final results of the communities. The proposed algorithm LPA-TS reduces the randomness in the process of node selection and label propagation; hence, we might obtain stable communities by LPA-TS. In addition, LPA-TS combines small scale communities with adjacent communities in the second phase of the algorithm to improve the quality of community detection. Compared with other classical community detection algorithms on some real networks and artificial networks, the proposed algorithm shows preferable performance on stability, NMI, ARI and modularity.
    Related Articles | Metrics
    An Approach for Storytelling by Correlating Events from Social Networks
    Li Yingying, Ma Shuai, Jiang Haoyi, Liu Zhe, Hu Chunming, Li Xiong
    Journal of Computer Research and Development    2018, 55 (9): 1972-1986.   DOI: 10.7544/issn1000-1239.2018.20180155
    Abstract846)   HTML11)    PDF (4887KB)(333)       Save
    Social networks, such as Twitter and Sina weibo, have become popular platforms to report the public event. They provide valuable data for us to monitor events and their evolution. However, informal words and fragmented texts make it challenging to extract descriptive information. Monitoring the event progression from fast accumulation of microblogs is also difficult. To this end, we monitor the event progression with a common topic from the social network. This can help us to gain an overview and a detailed documentation of the events. In this paper, we use three consecutive components to meet this end. First, we use a structure based approach to detect events from the microblog dataset. Second, we cluster the events by their topics based on their latent semantic information, and define each cluster as a story. Third, we use a graph based approach to generate a storyline for each story. The storyline is denoted by a directed acyclic graph (DAG) with a summary to express the progression of events in the story. The user experience evaluation indicates that this method can help us to monitor events and their progression by achieving improved accuracy and comprehension compared with the state of art methods.
    Related Articles | Metrics
    Exploration on Neural Information Retrieval Framework
    Guo Jiafeng, Fan Yixing
    Journal of Computer Research and Development    2018, 55 (9): 1987-1999.   DOI: 10.7544/issn1000-1239.2018.20180133
    Abstract1474)   HTML9)    PDF (3042KB)(803)       Save
    After decades of research, information retrieval technology has been significantly advanced and widely applied in our daily life. However, there is still a huge gap between modern search engines and true intelligent information accessing systems. In our opinion, an intelligent information accessing system should be able to crawl, read and understand the content of the big Web data, index and search the key semantic information, and reason, decide and generate the right results based on users’ information need. To develop such kind of systems, we need theoretical breakthrough on the search architecture and models. In recent years, to address the intelligent information accessing problem, we have conducted systematical research on neural information retrieval framework. We have achieved a few of original contributions on text representation, data indexing and relevance matching. However, there is still a long way in this direction and we will continue our exploration on neural information retrieval in the future.
    Related Articles | Metrics