ISSN 1000-1239 CN 11-1777/TP


    Please wait a minute...
    For Selected: Toggle Thumbnails
    Blockchain-Based Data Transparency: Issues and Challenges
    Meng Xiaofeng, Liu Lixin
    Journal of Computer Research and Development    2021, 58 (2): 237-252.   DOI: 10.7544/issn1000-1239.2021.20200017
    Abstract491)      PDF (1812KB)(499)       Save
    With the high-speed development of Internet of things, wearable devices and mobile communication technology, large-scale data continuously generate and converge to multiple data collectors, which influences people’s life in many ways. Meanwhile, it also causes more and more severe privacy leaks. Traditional privacy aware mechanisms such as differential privacy, encryption and anonymization are not enough to deal with the serious situation. What is more, the data convergence leads to data monopoly which hinders the realization of the big data value seriously. Besides, tampered data, single point failure in data quality management and so on may cause untrustworthy data-driven decision-making. How to use big data correctly has become an important issue. For those reasons, we propose the data transparency, aiming to provide solution for the correct use of big data. Blockchain originated from digital currency has the characteristics of decentralization, transparency and immutability, and it provides an accountable and secure solution for data transparency. In this paper, we first propose the definition and research dimension of the data transparency from the perspective of big data life cycle, and we also analyze and summary the methods to realize data transparency. Then, we summary the research progress of blockchain-based data transparency. Finally, we analyze the challenges that may arise in the process of blockchain-based data transparency.
    Related Articles | Metrics
    Fairness Research on Deep Learning
    Chen Jinyin, Chen Yipeng, Chen Yiming, Zheng Haibin, Ji Shouling, Shi Jie, Cheng Yao
    Journal of Computer Research and Development    2021, 58 (2): 264-280.   DOI: 10.7544/issn1000-1239.2021.20200758
    Abstract324)      PDF (1752KB)(265)       Save
    Deep learning is an important field of machine learning research, which is widely used in industry for its powerful feature extraction capabilities and advanced performance in many applications. However, due to the bias in training data labeling and model design, research shows that deep learning may aggravate human bias and discrimination in some applications, which results in unfairness during the decision-making process, thereby will cause negative impact to both individuals and socials. To improve the reliability of deep learning and promote its development in the field of fairness, we review the sources of bias in deep learning, debiasing methods for different types biases, fairness measure metrics for measuring the effect of debiasing, and current popular debiasing platforms, based on the existing research work. In the end we explore the open issues in existing fairness research field and future development trends.
    Related Articles | Metrics
    Survey on Automatic Text Summarization
    Li Jinpeng, Zhang Chuang, Chen Xiaojun, Hu Yue, Liao Pengcheng
    Journal of Computer Research and Development    2021, 58 (1): 1-21.   DOI: 10.7544/issn1000-1239.2021.20190785
    Abstract694)      PDF (1756KB)(829)       Save
    In recent years, the rapid development of Internet technology has greatly facilitated the daily life of human, and it is inevitable that massive information erupts in a blowout. How to quickly and effectively obtain the required information on the Internet is an urgent problem. The automatic text summarization technology can effectively alleviate this problem. As one of the most important fields in natural language processing and artificial intelligence, it can automatically produce a concise and coherent summary from a long text or text set through computer, in which the summary should accurately reflect the central themes of source text. In this paper, we expound the connotation of automatic summarization, review the development of automatic text summarization technique and introduce two main techniques in detail: extractive and abstractive summarization, including feature scoring, classification method, linear programming, submodular function, graph ranking, sequence labeling, heuristic algorithm, deep learning, etc. We also analyze the datasets and evaluation metrics that are commonly used in automatic summarization. Finally, the challenges ahead and the future trends of research and application have been predicted.
    Related Articles | Metrics
    Deep Neural Architecture Search: A Survey
    Meng Ziyao, Gu Xue, Liang Yanchun, Xu Dong, Wu Chunguo
    Journal of Computer Research and Development    2021, 58 (1): 22-33.   DOI: 10.7544/issn1000-1239.2021.20190851
    Abstract582)      PDF (1178KB)(680)       Save
    Deep learning has achieved excellent results on data tasks with multiple modalities such as images, speech, and text. However, designing networks manually for specific tasks is time-consuming and requires a certain level of expertise and design experience from the designer. In the face of today’s increasingly complex network architectures, relying on manual design alone increasingly becomes complex. For this reason, automatic architecture search of neural networks with the help of algorithms has become a hot research topic. The approach of neural architecture search involves three aspects: search space, search strategy, and performance evaluation strategy. The search strategy samples a network architecture in the search space, evaluates the network architecture by a performance evaluation strategy, and feed-back the results to the search strategy to guide it to select a better network architecture, and obtains the optimal network architecture through continuous iterations. In order to better sort out the methods of neural architecture search, we summarize the common methods in recent years from search space, search strategy and performance evaluation strategy, and analyze their strengths and weaknesses.
    Related Articles | Metrics
    Video Delivery over Named Data Networking: A Survey
    Hu Xiaoyan, Tong Zhongq, Xu Ke, Zhang Guoqiang, Zheng Shaoqi, Zhao Lixia, Cheng Guang, Gong Jian
    Journal of Computer Research and Development    2021, 58 (1): 116-136.   DOI: 10.7544/issn1000-1239.2021.20190697
    Abstract147)      PDF (1263KB)(89)       Save
    The Internet has developed into a network dominated by content delivery services such as delivering live and on-demand videos. There are some problems in traditional IP network in terms of supporting video delivery, such as the complexity and high overhead of the deployment of multicast, the disability to effectively utilize multipath transmission, the poor support for mobility and so on. Named data networking (NDN), a promising future Internet architecture, intrinsically supports in-network caching and multipath transmission. Consumers actively use interest message to request data packet from producer, and this consumer-driven communication model enables NDN to naturally support the mobility of consumers. These features offer the potential for NDN to efficiently deliver videos. This paper first introduces the background of video delivery and NDN, and then elaborates some schemes that take the advantages of NDN to deliver video: firstly, how do the strategies in NDN improve video bit rate; secondly, how do the strategies in NDN improve video playback stability; thirdly, how do the strategies in NDN protect video copyright and privacy; finally, how do the strategies in NDN transfer new types of video. According to the analysis of these existing schemes and the comparison of their performance over IP and NDN, the challenges of delivering videos over NDN are finally pointed out.
    Related Articles | Metrics
    Resource Management of Service Function Chain in NFV Enabled Network: A Survey
    Zu Jiachen, Hu Guyu, Yan Jiajie, Li Shiji
    Journal of Computer Research and Development    2021, 58 (1): 137-152.   DOI: 10.7544/issn1000-1239.2021.20190823
    Abstract202)      PDF (2139KB)(242)       Save
    With the emergence of new network technologies such as cloud computing, software-defined network (SDN) and network function virtualization (NFV), the future network’s management is supposed to become virtual and intelligent. NFV provides an approach to realize service functions based on the virtualization technology, and it adopts general servers to substitute the dedicated middlebox in traditional network, which is able to greatly reduce the capital expenditure (CAPEX) and the operating expense (OPEX) of the telecom service provider (TSP). NFV can also improve flexibility and scalability in the management of network services. Since the end-to-end network services are usually composed of different service functions, it is an important research topic to adopt virtualization technology to build service function chain (SFC) and reasonably allocate and schedule resources. In this paper, based on the background of NFV technology, we introduce the infrastructure, technical basis, and application scenarios of SFC in the NFV enabled network. Afterward, we mainly focus on the different stages of SFC orchestration: SFC composition, SFC placement, SFC scheduling, and SFC adaptive scaling. The correlated existing theoretical research is summarized. Finally, in view of the existing problems, some solutions are proposed and the future research directions are prospected.
    Related Articles | Metrics
    Research Advances on Knowledge Tracing Models in Educational Big Data
    Hu Xuegang, Liu Fei, Bu Chenyang
    Journal of Computer Research and Development    2020, 57 (12): 2523-2546.   DOI: 10.7544/issn1000-1239.2020.20190767
    Abstract735)   HTML6)    PDF (2358KB)(556)       Save
    With the in-depth advancement of informational education and the rapid development of online education, a large amount of fragmented educational data are generated during the learning process of students. How to mine and analyze these educational big data has become an urgent problem in the education and the knowledge engineering with big data fields. As for the dynamic education data, knowledge tracing models trace the cognitive status of students over time by analyzing the students’ exercising data generated in the learning process, so as to predict the exercising performance of students in the future time. In this paper, knowledge tracing models in educational big data are reviewed, analyzed, and discussed. Firstly, knowledge tracing models are introduced in detail from the perspective of their principles, steps, and model variants, including two mainstream knowledge tracing models based on Bayesian methods and deep learning methods. Then, the application scenarios of knowledge tracing models are explained from five aspects: student performance prediction, cognitive state assessment, psychological factor analysis, exercise sequence, and programming practice. The strengths and weaknesses in Bayesian knowledge tracing models and Deep Knowledge Tracing models are discussed through the two classic models BKT and DKT. Finally, some future directions of knowledge tracing models are given.
    Related Articles | Metrics
    Review of Automatic Image Annotation Technology
    Ma Yanchun, Liu Yongjian, Xie Qing, Xiong Shengwu, Tang Lingli
    Journal of Computer Research and Development    2020, 57 (11): 2348-2374.   DOI: 10.7544/issn1000-1239.2020.20190793
    Abstract685)   HTML16)    PDF (1358KB)(505)       Save
    As one of the most effective ways to reduce the “semantic gap” between image data and its content, automatic image annotation (AIA) technology has shown its great significance to help people understand image contents and retrieve the target information from the massive image data. This paper summarizes the general framework of AIA models by investigating the literatures about image annotation in recent 20 years, and analyzes the general problems to solve in AIA problems by combining the framework with various specific works. In this paper, the main methods used in various AIA models are classified into 9 types: correlation model, hidden Markov model, topic model, matrix factorization model, neighbor-based model, SVM-based model, graph-based model, CCA (KCCA) model and deep learning model. For each type of image annotation model, this paper provides a detailed study and analysis in terms of “basic principle introduction-specific model differences-model summary”. In addition, this paper summarizes some commonly used datasets and evaluation indexes, and compares the performance of some important image annotation models with related analysis on the advantages and disadvantages of various types of AIA models. Finally, some open problems and research directions in the field of image annotation are proposed and suggested.
    Related Articles | Metrics
    Survey on Geometric Unfolding, Folding Algorithms and Applications
    Sun Xiaopeng, Liu Shihan, Wang Zhenyan, Li Jiaojiao
    Journal of Computer Research and Development    2020, 57 (11): 2389-2403.   DOI: 10.7544/issn1000-1239.2020.20200126
    Abstract240)   HTML7)    PDF (1589KB)(235)       Save
    Unfolding and folding problem is a popular research topic in computer graphics, and has a wide range of applications, such as industrial manufacturing, architectural design, medical treatment, and aviation technology. In this survey, we review the basic concepts of unfolding and folding problem, introduce the research and application in four fields: robot design, computer animation, deep learning and others. We discuss the research work of unfolding and folding problem in detail. First, according to the different degrees of unfolding, we summarize research progress and typical algorithm ideas from two aspects: full unfolding and approximate unfolding. Full unfolding is to unfold 3D objects into 2D space without overlapping and deformation. However, most objects cannot be directly unfolded, and only an approximately unfolded structure can be solved. Approximate unfolding is a non-overlapping and deformed process, which is unfolded into the plane domain by mapping. How to find the smallest deformation is the key to approximate unfolding. Second, according to the different folding forms, the folding problem is divided into two types: Origami and Kirigami. We divide Origami into rigid folding and curved folding according to the different forms of crease, such as straight crease and curved crease. Kirigami is a special folding method that combines cutting and folding technology, which drives folding by the elastic force or other external forces generated by cutting. Here, we mainly consider the technology or algorithm of using Kirigami technology to construct auxetic structures. In addition, in order to compare the advantages and disadvantages of the algorithm, we summarize the commonly used algorithm indicators of unfolding and folding algorithm. Then, we evaluate the typical algorithm in recent years, and analyze advantages and disadvantages. Finally, we summarize and propose the development trend of unfolding and folding, including algorithm accuracy and robustness, fold volumetric objects, self-driven process and intelligent application of Kirigami technology.
    Related Articles | Metrics
    Survey on Data Updating in Erasure-Coded Storage Systems
    Zhang Yao, Chu Jiajia, Weng Chuliang
    Journal of Computer Research and Development    2020, 57 (11): 2419-2431.   DOI: 10.7544/issn1000-1239.2020.20190675
    Abstract212)   HTML7)    PDF (1644KB)(135)       Save
    In a distributed storage system, node failure has become a normal state. In order to ensure high availability of data, the system usually adopts data redundancy. At present, there are mainly two kinds of redundancy mechanisms. One is multiple replications, and the other is erasure coding. With the increasing amount of data, the benefits of the multi-copy mechanism are getting lower and lower, and people are turning their attention to erasure codes with higher storage efficiency. However, the complicated rules of the erasure coding itself cause the overhead of the read, write, and update operations of the distributed storage systems using the erasure coding to be larger than that of the multiple copies. Therefore, erasure coding is usually used for cold data or warm data storage. Hot data, which requires frequent access and update, is still stored in multiple copies. This paper focuses on the data update in erasure-coded storage systems, summarizes the current optimization work related to erasure coding update from the aspects of hard disk I/O, network transmission and system optimization, makes a comparative analysis on the update performance of representative coding schemes at present, and finally looks forward to the future research trends. Through analysis, it is concluded that the current erasure coding update schemes still cannot obtain the update performance similar to that of multiple copies. How to optimize the erasure-coded storage system in the context of erasure coding update rules and system architecture, so that it can replace the multi-copy mechanism under the hot data scenario, and reducing the hot data storage overhead is still a problem worthy of further study in the future.
    Related Articles | Metrics
    Principle and Research Progress of Quantum Computation and Quantum Cryptography
    Wang Yongli, Xu Qiuliang
    Journal of Computer Research and Development    2020, 57 (10): 2015-2026.   DOI: 10.7544/issn1000-1239.2020.20200615
    Abstract1308)   HTML24)    PDF (967KB)(1307)       Save
    Quantum computation and quantum cryptography are based on principles of quantum mechanics. In 1984, Bennett and Brassard proposed the first quantum key distribution protocol called BB84, which started the study of quantum cryptography. Since then, a great deal of work has been carried out in various fields such as quantum encryption and quantum signature. In 1994, Shor designed the first practical quantum algorithm which can factor large integers in polynomial time. Shor’s algorithm used Quantum Fourier Transform, which is the kernel of most quantum algorithms. In 1996, Grover designed a new algorithm which can search the unstructured data to get the required result in the time of approximately the square root of the total account of the data. Shor’s algorithm and Grover’s algorithm not only embody the advantages of quantum computing, but also pose a threat to the traditional cryptography based on mathematical difficulties such as RSA. After half a century’s development, quantum computing and quantum cryptography have achieved fruitful results in theory and practice. In this paper, we summarize the contents from the perspectives of the mathematical framework of quantum mechanics, basic concepts and principles, basic ideas of quantum computing, research progress and main ideas of quantum cryptography, etc.
    Related Articles | Metrics
    Research Advances on Privacy Preserving in Edge Computing
    Zhou Jun, Shen Huajie, Lin Zhongyun, Cao Zhenfu, Dong Xiaolei
    Journal of Computer Research and Development    2020, 57 (10): 2027-2051.   DOI: 10.7544/issn1000-1239.2020.20200614
    Abstract940)   HTML12)    PDF (3203KB)(763)       Save
    The wide exploitation of the theory of mobile communication and big data has enabled the flourishment of the outsourced system, where resource-constrained local users delegate batch of files and time-consuming evaluation tasks to the cloud server for outsourced storage and outsourced computation. Unfortunately, one single cloud server tends to become the target of comprise attack and bring about huge delay in response to the multi-user and multi-task setting where large quantity of inputs and outputs are respectively fed to and derived from the function evaluation, owing to its long distance from local users. To address this bottleneck of outsourced system, edge computing emerges that several edge nodes located between the cloud server and users collaborate to fulfill the tasks of outsourced storage and outsourced computation, meeting the real-time requirement but incurring new challenging issues of security and privacy-preserving. This paper firstly introduces the unique network architecture and security model of edge computing. Then, the state-of-the-art works in the field of privacy preserving of edge computing are elaborated, classified, and summarized based on the cryptographic techniques of data perturbation, fully homomorphic encryption, secure multiparty computation, fully homomorphic data encapsulation mechanism and verifiability and accountability in the following three phases: privacy-preserving data aggregation, privacy-preserving outsourced computation and their applications including private set intersection, privacy-preserving machine learning, privacy-preserving image processing, biometric authentication and secure encrypted search. Finally, several open research problems in privacy-preserving edge computing are discussed with convincing solutions, which casts light on its development and applications in the future.
    Related Articles | Metrics
    Overview of Threat Intelligence Sharing and Exchange in Cybersecurity
    Lin Yue, Liu Peng, Wang He, Wang Wenjie, Zhang Yuqing
    Journal of Computer Research and Development    2020, 57 (10): 2052-2065.   DOI: 10.7544/issn1000-1239.2020.20200616
    Abstract654)   HTML15)    PDF (1049KB)(641)       Save
    The emerging threats in cyberspace are endangering the interests of individuals, organizations and governments with complex and changeable attack methods. When traditional network security defense methods are not strong enough, the threat intelligence sharing and exchange mechanism has brought hope to the protection of cyberspace security. Cybersecurity threat intelligence is a collection of information that can cause potential harm and direct harm to organizations and institutions. This information can help organizations and institutions study and judge the cybersecurity threats they face, and make decisions and defenses accordingly. The exchange and sharing of threat intelligence can maximize the value of threat intelligence, reduce the cost of intelligence search and allieviate the problem of information islands, thereby improving the threat detection and emergency response capabilities of all parties involved in the sharing. This article first introduces the concept of cyber security threat intelligence and mainstream threat intelligence sharing norms; secondly, it investigates the literature on threat intelligence sharing and exchange at home and abroad in the past 10 years, and analyzes and summarizes the current situation and development trend of threat intelligence sharing and exchange. The article focuses on in-depth analysis from three perspectives of sharing models and mechanisms, the distribution of benefits of the exchange mechanism, and the privacy protection of shared data. The problems in the three parts and related solutions are pointed out, and the advantages and disadvantages of each solution are analyzed and discussed. Finally, the future research trend and direction of threat intelligence sharing and exchange are prospected.
    Related Articles | Metrics
    Security Issues and Privacy Preserving in Machine Learning
    Wei Lifei, Chen Congcong, Zhang Lei, Li Mengsi, Chen Yujiao, Wang Qin
    Journal of Computer Research and Development    2020, 57 (10): 2066-2085.   DOI: 10.7544/issn1000-1239.2020.20200426
    Abstract1418)   HTML23)    PDF (2361KB)(1197)       Save
    In recent years, machine learning has developed rapidly, and it is widely used in the aspects of work and life, which brings not only convenience but also great security risks. The security and privacy issues have become a stumbling block in the development of machine learning. The training and inference of the machine learning model are based on a large amount of data, which always contains some sensitive information. With the frequent occurrence of data privacy leakage events and the aggravation of the leakage scale annually, how to make sure the security and privacy of data has attracted the attention of the researchers from academy and industry. In this paper we introduce some fundamental concepts such as the adversary model in the privacy preserving of machine learning and summarize the common security threats and privacy threats in the training and inference phase of machine learning, such as privacy leakage of training data, poisoning attack, adversarial attack, privacy attack, etc. Subsequently, we introduce the common security protecting and privacy preserving methods, especially focusing on homomorphic encryption, secure multi-party computation, differential privacy, etc. and compare the typical schemes and applicable scenarios of the three technologies. At the end, the future development trend and research direction of machine learning privacy preserving are prospected.
    Related Articles | Metrics
    Internet Data Transfer Protocol QUIC: A Survey
    Li Xuebing, Chen Yang, Zhou Mengying, Wang Xin
    Journal of Computer Research and Development    2020, 57 (9): 1864-1876.   DOI: 10.7544/issn1000-1239.2020.20190693
    Abstract960)   HTML30)    PDF (929KB)(650)       Save
    QUIC is an Internet data transfer protocol proposed by Google as an alternative for TCP (transmission control protocol). Compared with TCP, QUIC introduces lots of new features to make it theoretically outperform TCP in many fields. For example, it supports multiplexing to solve the problem of head-of-line blocking, introduces 0-RTT handshake to reduce handshake latency, and supports connection migration to be mobility-friendly. However, QUIC’s performance in the real world may not be as good as expected, because network environments and network devices are diverse and the protocol’s security is challenged by potential attackers. Therefore, evaluating QUIC’s impact on existing network services is quite important. This paper carries out a comprehensive survey of QUIC. We introduce the development history and the main characteristics of QUIC firstly. Secondly, taking the two most widely used application scenarios: Web browsing and video streaming as examples, we introduce and summarize domestic and international research analysis on the data transmission performance of QUIC under different network environments. Thirdly, we enumerate existing QUIC-enhancement work from the aspects of protocol design and system design. Fourthly, we summarize existing work on the security analysis on QUIC. We enumerate the security issues that are currently recognized by the academic community, as well as the researchers’ efforts to address these issues. Lastly, we come up with several potential improvements on existing research outcomes and look forward to new research topics and challenges brought by QUIC.
    Related Articles | Metrics
    Interpretation and Understanding in Machine Learning
    Chen Kerui, Meng Xiaofeng
    Journal of Computer Research and Development    2020, 57 (9): 1971-1986.   DOI: 10.7544/issn1000-1239.2020.20190456
    Abstract1475)   HTML54)    PDF (1315KB)(1352)       Save
    In recent years, machine learning has developed rapidly, especially in the deep learning, where remarkable achievements are obtained in image, voice, natural language processing and other fields. The expressive ability of machine learning algorithm has been greatly improved; however, with the increase of model complexity, the interpretability of computer learning algorithm has deteriorated. So far, the interpretability of machine learning remains as a challenge. The trained models via algorithms are regarded as black boxes, which seriously hamper the use of machine learning in certain fields, such as medicine, finance and so on. Presently, only a few works emphasis on the interpretability of machine learning. Therefore, this paper aims to classify, analyze and compare the existing interpretable methods; on the one hand, it expounds the definition and measurement of interpretability, while on the other hand, for the different interpretable objects, it summarizes and analyses various interpretable techniques of machine learning from three aspects: model understanding, prediction result interpretation and mimic model understanding. Moreover, the paper also discusses the challenges and opportunities faced by machine learning interpretable methods and the possible development direction in the future. The proposed interpretation methods should also be useful for putting many research open questions in perspective.
    Related Articles | Metrics
    Review of Entity Relation Extraction Methods
    Li Dongmei, Zhang Yang, Li Dongyuan, Lin Danqiong
    Journal of Computer Research and Development    2020, 57 (7): 1424-1448.   DOI: 10.7544/issn1000-1239.2020.20190358
    Abstract1679)   HTML44)    PDF (1404KB)(1291)       Save
    There is a phenomenon that information extraction has long been concerned by a lot of research works in the field of natural language processing. Information extraction mainly includes three sub-tasks: entity extraction, relation extraction and event extraction, among which relation extraction is the core mission and a great significant part of information extraction. Furthermore, the main goal of entity relation extraction is to identify and determine the specific relation between entity pairs from plenty of natural language texts, which provides fundamental support for intelligent retrieval, semantic analysis, etc, and improves both search efficiency and the automatic construction of the knowledge base. Then, we briefly expound the development of entity relation extraction and introduce several tools and evaluation systems of relation extraction in both Chinese and English. In addition, four main methods of entity relation extraction are mentioned in this paper, including traditional relation extraction methods, and other three methods respectively based on traditional machine learning, deep learning and open domain. What is more important is that we summarize the mainstream research methods and corresponding representative results in different historical stages, and conduct contrastive analysis concerning different entity relation extraction methods. In the end, we forecast the contents and trend of future research.
    Related Articles | Metrics
    An Energy Consumption Optimization and Evaluation for Hybrid Cache Based on Reinforcement Learning
    Fan Hao, Xu Guangping, Xue Yanbing, Gao Zan, Zhang Hua
    Journal of Computer Research and Development    2020, 57 (6): 1125-1139.   DOI: 10.7544/issn1000-1239.2020.20200010
    Abstract765)   HTML20)    PDF (3887KB)(505)       Save
    Emerging non-volatile memory STT-RAM has the characteristics of low leakage power, high density, fast read speed, and high write energy. Meanwhile, SRAM has the characteristics of high leakage power, low density, fast read and write speed, low write energy, etc. The hybrid cache of SRAM and STT-RAM fully utilizes the respective advantages of both memory medias, providing lower leakage power and higher cell density than SRAM, higher write speed and lower write energy than STT-RAM. The architecture of hybrid cache mainly achieves both of benefits by putting write-intensive data into SRAM and read-intensive data into STT-RAM. Therefore, how to identify and allocate read-write-intensive data is the key challenge for the hybrid cache design. This paper proposes a cache management method based on the reinforcement learning that uses the write intensity and reuse information of cache access requests to design a cache allocation policy and optimize energy consumption. The key idea is to use the reinforcement learning algorithm to get the weight for the set allocating to SRAM or STT-RAM by learning from the energy consumption of cache line sets. The algorithm allocates a cache line in a set to the region with greater weight. Evaluations show that our proposed policy reduces the average energy consumption by 16.9%(9.7%) in a single-core (quad-core) system compared with the previous policies.
    Related Articles | Metrics
    Optimizing Winograd-Based Fast Convolution Algorithm on Phytium Multi-Core CPUs
    Wang Qinglin, Li Dongsheng, Mei Songzhu, Lai Zhiquan, Dou Yong
    Journal of Computer Research and Development    2020, 57 (6): 1140-1151.   DOI: 10.7544/issn1000-1239.2020.20200107
    Abstract554)   HTML7)    PDF (2411KB)(343)       Save
    Convolutional neural networks (CNNs) have been extensively used in artificial intelligence fields such as computer vision and natural language processing. Winograd-based fast convolution algorithms can effectively reduce the computational complexity of convolution operations in CNNs so that they have attracted great attention. With the application of Phytium multi-core CPUs independently developed by the National University of Defense Technology in artificial intelligence fields, there is strong demand of high-performance convolution primitives for Phytium multi-core CPUs. This paper proposes a new high-performance parallel Winograd-based fast convolution algorithm after studying architecture characteristics of Phytium multi-core CPUs and computing characteristics of Winograd-based fast convolution algorithms. The new parallel algorithm does not rely on general matrix multiplication routines, and consists of four stages: kernels transformation, input feature maps transformation, element-wise multiplication, and output feature maps inverse transformation. The data movements in all four stages have been collaboratively optimized to improve memory access performance of the algorithm. The custom data layouts, multi-level parallel data transformation algorithms and multi-level parallel matrix multiplication algorithm have also been proposed to support the optimization above efficiently. The algorithm is tested on two Phytium multi-core CPUs. Compared with Winograd-based fast convolution implementations in ARM Computer Library (ACL) and NNPACK, the algorithm can achieve speedup of 1.05~16.11 times and 1.66~16.90 times, respectively. The application of the algorithm in the open source framework Mxnet improves the forward-propagation performance of the VGG16 network by 3.01~6.79 times.
    Related Articles | Metrics
    Efficient Optimization of Graph Computing on High-Throughput Computer
    Zhang Chenglong, Cao Huawei, Wang Guobo, Hao Qinfen, Zhang Yang, Ye Xiaochun, Fan Dongrui
    Journal of Computer Research and Development    2020, 57 (6): 1152-1163.   DOI: 10.7544/issn1000-1239.2020.20200115
    Abstract553)   HTML7)    PDF (1876KB)(313)       Save
    With the rapid development of computing technology, the scale of graph increases explosively and large-scale graph computing has been the focus in recent years. Breadth first search (BFS) is a classic algorithm to solve graph traverse problem. It is the main kernel of Graph500 benchmark that evaluates the performance of supercomputers and servers in terms of data-intensive applications. High-throughput computer (HTC) adopts ARM-based many-core architecture, which has the characteristics of high concurrency, strong real-time, low-power consumption. The optimization of BFS algorithm has made a series of progress on single-node systems. In this paper, we first introduce parallel BFS algorithm and existing optimizations. Then we propose two optimization techniques for HTC to improve the efficiency of data access and data locality. We systematically evaluate the performance of BFS algorithm on HTC. For the Kronecker graph with 2scale=230whose vertices are 230 and edges are 234, the average performance on HTC is 24.26 GTEPS and 1.18 times faster than the two-way x86 server. In terms of energy efficiency, the result on HTC is 181.04 MTEPS/W and rank 2nd place on the June 2019 Green Graph500 big data list. To our best knowledge, this is the first work that evaluates BFS performance on HTC platform. HTC is suitable for data intensive applications such as large-scale graph computing.
    Related Articles | Metrics