| 作成者 |
|
|
|
|
|
| 本文言語 |
|
| 出版者 |
|
|
|
| 発行日 |
|
| 収録物名 |
|
| 巻 |
|
| 出版タイプ |
|
| アクセス権 |
|
| 関連DOI |
|
|
|
| 関連URI |
|
|
|
| 関連情報 |
|
|
|
| 概要 |
A learning method using domains of attraction in three-layered neural net-works is proposed. The method is a combination of output error minimization learning with maximization learning of domains of ...attraction in one-layered perceptrons. To simplify the structure of the network, a successive learning technique is employed for hidden units. Domains of attraction in the network is derived by restricting output conditions at the hidden layer for training input data. A three-layered neural network is determined by this method using a training set which consists of satellite observation data and soil moisture data surveyed in some places. The estimation of soil moisture at all places corresponding to satellite data is carried out based on the derived domains of attraction in the network.続きを見る
|
Mendeley出力
trcs152
trcs152.ps