講演抄録/キーワード |
講演名 |
2016-11-24 15:10
Modified Nearest Feature Space Approach for High Dimensional Data Sets ○Yang Lang Chang・Yung-Hao Lai・Tzu-Wei Tseng・Jyh-Perng Fang(NTUT) SANE2016-65 |
抄録 |
(和) |
With the progress of remote sensing technology, the high volume of high dimensional data sets is increased rapidly. Thus, it is important to better analyze and particular process these huge datasets. Several studies have been investigated to the classification algorithm of high dimensional data sets, such as nearest feature space (NFS). NFS can provide more information to virtually enlarge the training samples set. However, NFS is difficult to categorize and easily cause misjudgments when test point is too close with the distribution of sample points. In this paper, we proposed a new method called modified nearest feature space (MNFS), which can analysis the coverage of the feature space (FS) used in NFS and limit the extensible range of each FS in line, to reduce the impact of the overlapping between each category. Experimental results prove that MNFS is better than NFS on classification accuracy. |
(英) |
With the progress of remote sensing technology, the high volume of high dimensional data sets is increased rapidly. Thus, it is important to better analyze and particular process these huge datasets. Several studies have been investigated to the classification algorithm of high dimensional data sets, such as nearest feature space (NFS). NFS can provide more information to virtually enlarge the training samples set. However, NFS is difficult to categorize and easily cause misjudgments when test point is too close with the distribution of sample points. In this paper, we proposed a new method called modified nearest feature space (MNFS), which can analysis the coverage of the feature space (FS) used in NFS and limit the extensible range of each FS in line, to reduce the impact of the overlapping between each category. Experimental results prove that MNFS is better than NFS on classification accuracy. |
キーワード |
(和) |
nearest feature space classifier / fisher criterion principle component analysis / / / / / / |
(英) |
nearest feature space classifier / fisher criterion principle component analysis / / / / / / |
文献情報 |
信学技報, vol. 116, no. 319, SANE2016-65, pp. 71-73, 2016年11月. |
資料番号 |
SANE2016-65 |
発行日 |
2016-11-17 (SANE) |
ISSN |
Print edition: ISSN 0913-5685 Online edition: ISSN 2432-6380 |
著作権に ついて |
技術研究報告に掲載された論文の著作権は電子情報通信学会に帰属します.(許諾番号:10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
PDFダウンロード |
SANE2016-65 |