A Contemporary and Comprehensive Survey on Streaming Tensor Decomposition

Tensor decomposition has been demonstrated to be successful in a wide range of applications, from neuroscience and wireless communications social networks. In an online setting, factorizing tensors derived multidimensional data streams is however nontrivial due several inherent problems real-time stream processing. recent years, many research efforts have dedicated developing techniques for decomposing such tensors, resulting significant advances streaming tensor or tracking. This topic emerging enriches the literature on decomposition, particularly analystics perspective. Thus, it imperative carry out overview tracking help researchers practitioners understand its developments achievements, summarise current trends advances, identify challenging problems. this article, we provide contemporary comprehensive survey different types techniques. We categorize state-of-the-art methods into three main groups: CP decompositions, Tucker decompositions under other formats (i.e., tensor-train, t-SVD, BTD). each group, further divide existing algorithms sub-categories based their optimization framework model architectures. Finally, present challenges, open problems, potential directions future.