|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2007年03 期
| 標(biāo)題 |
| 基于RBF神經(jīng)網(wǎng)絡(luò)的在線分類挖掘系統(tǒng)(16 卷) |
| 英文標(biāo)題 |
| On-line Classification System based on Dynamic RBF Neural Networks |
| 摘要 |
| 模式分類是RBF神經(jīng)網(wǎng)絡(luò)應(yīng)用的一個(gè)重要方面。在線環(huán)境中數(shù)據(jù)集是經(jīng)常變動(dòng)的，采用批量式學(xué)習(xí)算法（如OLS算法）訓(xùn)練RBF網(wǎng)絡(luò)會(huì)產(chǎn)生大量的重復(fù)訓(xùn)練，從而導(dǎo)致學(xué)習(xí)效率不高。為彌補(bǔ)這種不足，從梯度下降方法推導(dǎo)出一種增量式學(xué)習(xí)算法，用于在線環(huán)境中的RBF神經(jīng)網(wǎng)絡(luò)訓(xùn)練。最后將基 |
| 作者 |
| 新聞作者:段錄平1 ，周麗娟2，王 宇1 |
| 關(guān)鍵字 |
