2013年6月
Relationship between phase and amplitude generalization errors in complex- and real-valued feedforward neural networks
NEURAL COMPUTING & APPLICATIONS
- ,
- 巻
- 22
- 号
- 7-8
- 開始ページ
- 1357
- 終了ページ
- 1366
- 記述言語
- 英語
- 掲載種別
- 研究論文(学術雑誌)
- DOI
- 10.1007/s00521-012-0960-z
- 出版者・発行元
- SPRINGER
We compare the generalization characteristics of complex-valued and real-valued feedforward neural networks. We assume a task of function approximation with phase shift and/or amplitude change in signals having various coherence. Experiments demonstrate that complex-valued neural networks show smaller generalization error than real-valued networks having doubled input and output neurons in particular when the signals have high coherence, that is, high degree of wave nature. We also investigate the relationship between amplitude and phase errors. It is found in real-valued networks that abrupt change in amplitude is often accompanied by steep change in phase, which is a consequence of local minima in real-valued supervised learning.
- リンク情報
-
- DOI
- https://doi.org/10.1007/s00521-012-0960-z
- DBLP
- https://dblp.uni-trier.de/rec/journals/nca/HiroseY13
- Web of Science
- https://gateway.webofknowledge.com/gateway/Gateway.cgi?GWVersion=2&SrcAuth=JSTA_CEL&SrcApp=J_Gate_JST&DestLinkType=FullRecord&KeyUT=WOS:000319769300012&DestApp=WOS_CPL
- URL
- http://dblp.uni-trier.de/db/journals/nca/nca22.html#journals/nca/HiroseY13
- ID情報
-
- DOI : 10.1007/s00521-012-0960-z
- ISSN : 0941-0643
- eISSN : 1433-3058
- DBLP ID : journals/nca/HiroseY13
- Web of Science ID : WOS:000319769300012