A Barzilai-Borwein Gradient Algorithm for Spatio-Temporal Internet Traffic Data Completion via Tensor Triple Decomposition

作     者：Chen, Yannan Zhang, Xinzhen Qi, Liqun Xu, Yanwei 

作者机构：South China Normal Univ Sch Math Sci Guangzhou Peoples R China Tianjin Univ Sch Math Tianjin 300354 Peoples R China Huawei Technol Investment Co Ltd Labs 2012 Future Network Theory Lab Shatin Hong Kong Peoples R China Hangzhou Dianzi Univ Dept Math Sch Sci Hangzhou 310018 Peoples R China Hong Kong Polytech Univ Dept Appl Math Hung Hom Kowloon Hong Kong Peoples R China Huawei Theory Res Lab Kowloon Hong Kong Peoples R China 

出 版 物：《JOURNAL OF SCIENTIFIC COMPUTING》 (科学计算杂志)

年 卷 期：2021年第88卷第3期

页      面：65-65页

核心收录：

学科分类：08[工学] 0701[理学-数学] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：National Natural Science Foundation of China [11771405, 12071159] Guangdong Basic and Applied Basic Research Foundation [2020A1515010489] NSFC 

主　　题：Internet traffic tensor Tensor completion Triple tensor decomposition Optimization algorithm 

摘      要：With the coming of high-speed network and 5G era, internet traffic data is crucial for various network tasks such as traffic engineering, capacity planning and anomaly detection. To explore the natural spatio-temporal structure of network flow, we use the novel triple decomposition of tensors to establish an optimization model with the spatio-temporal regularization for completing the internet traffic data. A Barzilai-Borwein gradient algorithm is designed for solving the spatio-temporal internet traffic tensor completion problem. We prove the convergence of this algorithm and analyze its convergence rate with the tool of the Kurdyka-Lojasiewicz property. Numerical experiments on Abilene and GeANT datasets report that the proposed tensor completion method is effective.