2020 Fiscal Year Final Research Report
Knowledge extraction from large-scale sequence data by combining pattern mining and sparse modeling
Project/Area Number |
17H04694
|
Research Category |
Grant-in-Aid for Young Scientists (A)
|
Allocation Type | Single-year Grants |
Research Field |
Intelligent informatics
|
Research Institution | Nagoya Institute of Technology |
Principal Investigator |
Karasuyama Masayuki 名古屋工業大学, 工学(系)研究科(研究院), 准教授 (40628640)
|
Project Period (FY) |
2017-04-01 – 2021-03-31
|
Keywords | 機械学習 / 疎性モデリング / パターンマイニング / 構造データ |
Outline of Final Research Achievements |
Because of development of a variety of sensor devices and increase of the capacity of storage devices, a significance of data-driven approaches to extracting useful knowledge from accumulated data have been widely recognized. In this study, we consider identifying important substructures in sequence or more complicated graph data, which has a connected structure inside a data instance. Since combinatorially many possible substructures exist, we propose efficient pruning algorithms that removes unnecessary patterns. Our framework is based on an optimization theory, and thus, the optimality of the resulting identified substructures can be guaranteed.
|
Free Research Field |
機械学習
|
Academic Significance and Societal Importance of the Research Achievements |
機械学習技術の注目が高まるつれ,その解釈性は大きな社会的関心事となっている.本課題で扱うような系列やグラフは,時系列データや化合物などの科学データで単純な数値テーブルでは表現できないデータの表現方法として広く定着しているものである.そのため,多様化するデータ駆動解析において,本課題で扱ったような問題設定は今後ますます顕在化すると考えられる.一方で,組み合わせ的に爆発する部分構造に対し,最適性を保証しつつ学習する枠組みはほとんど研究がなく,本研究の独自性・意義を示すものである.
|