https://scholars.lib.ntu.edu.tw/handle/123456789/85658
標題: | 巢狀超矩形學習模式於水資源系統之研究 A Study of the Nested Hyper-Rectangles Learning Model for Water Resources Systems |
作者: | 張斐章 陳莉 Chang, Fi-John Chen, Li |
關鍵字: | 水資源;系統 | 公開日期: | 九月-1992 | 起(迄)頁: | 27-37 | 來源出版物: | 中國農業工程學報 | 摘要: | 預測及分類為水文學者重要的工作。巢狀超矩形學習模式是一種「以範例為基礎的學習」(Exemplar-based learning)模式,最主要的觀念是將過去發生的許多事件以一個事件為一個點(Point)的型式貯存的歐基里得(Euclidean)n維空間(En)中,將來如果有一新的事件(一個新的點)加入模式的時候,就可以計算出與原來樣本空間中最接近的點,進而達成分類或預測的效果。 此模式因可經由新加入之範例而動態調整「距離計量」中的各項參數以回饋系統,故其準確度隨著所訓練樣本的增加而愈正確,即本模式可具有學習的能力。 為說明此一模式,本研究先設定一簡單的數學函數並以此模式預估其理論值。之後,並實際應用於河川流量的分級預報,日雨量記錄的補遺,及年平均流量之延伸等工作,結果皆顯示其優越的預報及分類能力,值得進一步的研究及推展。 Prediction and categorization are important tasks of hydrologists. The nested hyper-rectangles learning model which is an examplarbased learning model is studied and applied to above tasks. The main idea of the model is "seeding" history data in Euclidean n-space,Eⁿ,as exemplars, then comparing new examples (data) to those seeding points, and finding the most similar example in memory. A dynamic adjustment for model's parameters in proposed with the "distance metric" which is used to determine the similarity, so the simulated system can be represented or predicted more and more accurate through the feedback of added examples. That is the model has ability to learn. In order to show the general characteristics of the model performances, a simple mathematical function is simulated by the model. It is then applied to three different hydrological systems. They are: (1) forecasting streamflow categorization; (2) estimating the missing record of daily precipitation; and (3) extending the annual streamflow. The results demonstrate the power of hyperrectangle learning model, and the model can be a very useful tool on hydrological system. |
URI: | http://ntur.lib.ntu.edu.tw//handle/246246/106323 |
顯示於: | 生物環境系統工程學系 |
檔案 | 描述 | 大小 | 格式 | |
---|---|---|---|---|
巢狀超矩型學習模式於水資源系統之研究.pdf | 677.68 kB | Adobe PDF | 檢視/開啟 |
在 IR 系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。