查詢結果分析
相關文獻
- 非監督類神經網路之迭代合併式學習法則
- An On-Line Learning Neural Network for Radar Signal Classification
- 臺灣地區銀行業顧客滿意度之研究--自組織映射類神經網路之應用
- 徑向基類神經網路之運動平臺動態系統判別
- 整合遙測特徵指標與自組織聚類神經網路模式於崩塌地萃取之研究--以莫拉克風災之阿里山溪集水區為例
- 結合群聚分析與人工智慧於颱風時雨量預測
- 以共用臨域類神經網路架構土石流預警系統之研究
- Neural Network Procedures for Taguchi's Dynamic Problems
- A Fast and Efficient Competitive Learning Design Algorithm Based on Weight Vector Training in Transform Domain
- 專家系統振動訊號圖型判別之研究
頁籤選單縮合
| 題 名 | 非監督類神經網路之迭代合併式學習法則=A New Unsupervised Learning Algorithm Based on Iteratively Merging |
|---|---|
| 作 者 | 張志吉; 徐學群; 李岳軍; 李靖海; | 書刊名 | 中正嶺學報 |
| 卷 期 | 25:2 1997.01[民86.01] |
| 頁 次 | 頁195-206 |
| 分類號 | 593 |
| 關鍵詞 | 非監督式學習; 類神經網路; 聚類; 解交錯; Unsupervised learning; Neural network; Clustering; Deinterleaving; |
| 語 文 | 中文(Chinese) |
| 中文摘要 | 在學習過程中機動增加網路節點數目的非監督式類神經網路(如ART[1,2]),其 主要缺點為分類的結果與樣本的學習次序有著密切的關係。為改正此缺點,本文提出位於密 度較高區域的樣本能優先學習的觀念,並由此觀念發展出一個迭代合併式學習法則。在迭代 學習的過程中,會逐步地找出具有較高密度的次聚類 (sub-cluster)。然以這些次聚類的中 心作為下次迭代學習的樣本,因而使具有較高密度的次聚類逐步地將附近較小的次聚類合併 而產生正確的分類結果。本文以雷達脈波解交錯 (deinterleaving) 的分類問題來作驗證。 分類的結果於文中將詳加討論,並與機動增加節點數目的非監督式學習法則之分類結果作一 比較。 |
| 英文摘要 | The defect of the unsupervised neural networks which dynamically add nodes for clusters during learning (such as ART[,2]) is that clustering results highly depend on the sequence of input samples. To improve this defect, we think that the region with higher density of samples is more possible to be the center of a cluster. An iteratively merging learning algorithm is therefore developed based on this idea. In this learning process, sub-clusters with higher density will be found out and act as centers of the next learning iteration. By doing so, more and more sub-clusters with lower densities will be merged into neighbor sub-clusters with higher densities. An algorithm is evaluated by radar signal deinterleaving problem in this paper. Some results are discussed and compared with other unsupervised algorithm. |
本系統中英文摘要資訊取自各篇刊載內容。