講演抄録/キーワード |
講演名 |
2013-08-29 15:40
Conjugate Gradient Method For Orthogonal Matrix Optimization Using Cayley Transform ○Gen Hori(Asia Univ.) SIP2013-72 |
抄録 |
(和) |
Matrix optimization problems required by several applications in signal and image processing appear with orthogonality constraints. It has been shown that the natural gradient performs better than the conventional gradient in orthogonal matrix optimization using the Cayley transform. However, to develop practical orthogonal matrix optimization, it is desirable to apply faster algorithms such as the variable metric method and the conjugate gradient method. The present study applies those methods to orthogonal matrix optimization with the Cayley transform and proposes to replace the conventional gradient terms appeared in the update rules of those methods by the natural gradient. A numerical simulation using joint diagonalization problem shows that the replacement considerably improves the performance of those methods. |
(英) |
(Not available yet) |
キーワード |
(和) |
Orthogonal matrix optimization / Joint diagonalization / Natural gradient / Cayley transform / / / / |
(英) |
/ / / / / / / |
文献情報 |
信学技報, vol. 113, no. 191, SIP2013-72, pp. 31-34, 2013年8月. |
資料番号 |
SIP2013-72 |
発行日 |
2013-08-22 (SIP) |
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 |
SIP2013-72 |