Model Selection for Tensor Factorization and its Applications for Big Data Analysis

2017 Fiscal Year Final Research Report

• Project/Area Number: 15K16067
• Research Category: Grant-in-Aid for Young Scientists (B)
• Allocation Type: Multi-year Fund
• Research Field: Soft computing
• Research Institution: Nagoya Institute of Technology
• Principal Investigator: Yokota Tatsuya 名古屋工業大学, 工学(系)研究科(研究院), 助教 (80733964)
• Project Period (FY): 2015-04-01 – 2018-03-31
• Keywords: テンソル分解 / モデル選択 / 情報量基準 / テンソル補完 / テンソル核ノルム / テンソル総変動

Outline of Final Research Achievements

Tensor is a name of multi-dimensional array including vectors and matrices. For analyzing tensors, there are many approaches such as tensor factorization and tensor networks. However, model selection (i.e., rank estimation) is a critical issue of tensor factorization and its applications. In this study, we tackled the problem of model selection in tensor analysis while developing many algorithms for tensor rank estimation, noise reduction, completion, and super-resolution. Totally, we published four journal papers including arXiv and fourteens conference presentations including domestic and international conferences. It includes two highly impact journal papers published in the IEEE Transactions on Signal Processing, and two highly impact international conference papers accepted for the IEEE Conference on Computer Vision and Pattern Recognition.

Free Research Field

テンソル信号処理