2011 Fiscal Year Final Research Report
A Study on Data-Dependent Complexities
| Project/Area Number |
21500128
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 一般 |
|---|
| Research Field |
Intelligent informatics
|
|---|
| Research Institution | Hokkaido University |
|---|
Principal Investigator |
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
KUDO Mineichi 北海道大学, 大学院・情報科学研究科, 教授 (60205101)
TOYAMA Jun 北海道大学, 大学院・情報科学研究科, 助教 (60197960)
|
|---|
| Project Period (FY) |
2009 – 2011
|
| Keywords | 知識発見とデータマイニング / データ依存 / 問題の複雑さ / 機械学習 / シーケンス / 超矩形 / アルゴリズム |
|---|
| Research Abstract |
As a data-dependent complexity measure, we proposed data-dependent VC dimension of Sperner family, and in hyper-rectangle subclass problem, we developed a fast algorithm for datasets with small such VC dimension. We also empirically showed efficiency of the algorithm using real data. Furthermore, as a complexity measure of contiguous repetition structure of a string, we proposed the size of a minimum repetition representation string (MRRS) for a given string. We developed a fast algorithm for constructing an MRRS and analyzed the repetition structure of DNA sequences using the algorithm.
|
