We use cookies to improve your experience with our site.
Jia-Feng Guo, Yi-Qing Zhou. Preface[J]. Journal of Computer Science and Technology, 2022, 37(4): 741-742. DOI: 10.1007/s11390-022-0004-9
Citation: Jia-Feng Guo, Yi-Qing Zhou. Preface[J]. Journal of Computer Science and Technology, 2022, 37(4): 741-742. DOI: 10.1007/s11390-022-0004-9
  • Xia Peisu Young Scholars Forum 2021 is intended to promote thecommunication between young scholars world wide and exchange novelresearch ideas and methods in the frontier of computer science. Theforum is named after Prof. Xia Peisu, an academician of the ChineseAcademy of Sciences, who made a great contribution to the development ofcomputational technologies and participated in building the firstcomputer in China. Prof. Xia Peisu has also founded the first Englishcomputer journal of China, i.e., Journal of Computer Science andTechnology (JCST), in 1986.
    The main topic of Xia Peisu Young Scholars Forum 2021 is “Network,Data and AI”, which aims to discuss the opportunities and challenges indeveloping the big data, artificial intelligence and next-generationnetwork technologies. With the development of information technologies,we are entering an age of intelligence where network, data and AI havebecome indispensable elements which are intertwined closely and havehuge impact on the human society. The intensive discussions on the threetopics in this forum help promote the cross-disciplinary researchbetween these fields. After the forum, we invited the participants tosubmit their work to JCST. After two rounds of peer-review, eight paperswere selected for the Special Section of Xia Peisu Young Scholars Forum2021.
    In this special section, there are three papers on the jointoptimization of applications and networking. Traditionally, applicationstreat networks as a black box, which mainly provides informationtransport services. Meanwhile, networking optimizations focus on “lowerlatency and higher throughput” for decades. However, networking's realand only mission should be making applications better, and networkingschemes should be jointly optimized with applications. Being aware ofoccasional packet corruptions' detrimental effects to RDMA-enabledapplications, Gao et al.'s work strives to shrinkthe flow completion time by orders of magnitude. By understandingapplication semantics from passing-through packets, Dong etal.'s work could improve data-parallel jobs' performance byover 10%. Wang et al.'s workintelligently places energy harvesting nodes that can significantlyprolong the lifetime of a sensor service.
    Another two papers are related to distributive networking, which hasattracted much attention in recent years. Shi etal.'s work focuses on the performance assessment ofdecentralized clouds and proposes a robust assessment solution RODE.Experiments show that RODE can accurately monitor the performance ofcloud providers. Meanwhile, considering distributive platforms,iterative algorithms are promising to analyze large scale data. Yu et al.'s work designs an efficient executionmanager Aiter-R, which can be integrated into existing delta-basediterative processing to achieve maximum efficiency, with the proposedgroup-based iterative execution approach. Experimental results show thatAiter-R outperforms state-of-the-art solutions.
    Besides, there are three papers related to the topics of big data andartificial intelligence, including information retrieval, recommendationand text summarization. Wu et al.'s work focuseson the document ranking in information retrieval and investigates howusers' information gain accumulates both within a document and across aquery session. The proposed model PCGM, which incorporates thedocument-level and query-level passage cumulative gain, outperformsmultiple advanced ranking baselines and the predicted results are highlyconsistent with users' preferences. Jiang et al.'swork aims to improve the friend recommendation with fine-grainedevolving interests and proposes an LPRF-F framework which explores thelearning interest tags and time features to predict the user interests.Extensive experiments validate the effectiveness of this work as well asthe effects of social influence and cross-domain interest. Jiang et al.'s another paper strives to better evaluatethe performance of abstractive summarization models by consideringngram-based semantic information. Empirical results demonstrate theproposed evaluation metrics are well correlated with humanjudgements.
    Five papers of this special section are published in Vol.37, No.4,2022, and the other three, i.e., the papers of Shi etal. and Jiang et al., will be includedin Vol.37, No.5, 2022.
    We hope that readers will enjoy this special section. We are gratefulto all the paper authors and reviewers for their valuablecontributions.
  • loading



    DownLoad:  Full-Size Img  PowerPoint