查詢結果分析
來源資料
頁籤選單縮合
題 名 | 類神經網路解碼訓練法則在字型辨認上之應用=The Application of the Neural Networks Based Decode Algorithm on Character Recognition |
---|---|
作 者 | 李智勇; 樓康寧; | 書刊名 | 吳鳳學報 |
卷 期 | 5 1997.06[民86.06] |
頁 次 | 頁152-162 |
分類號 | 312.13 |
關鍵詞 | 解碼訓練法則; 字型辨認; 圖型辨識; 雙極性加權值; Decode algorithm; Character recognition; Pattern recognition; Bipolar weight; |
語 文 | 中文(Chinese) |
中文摘要 | 使用類神經網路來建立圖型及字型辨認的模式很多,但在現有模式中,卻存在著 訓練次數,參數調整及穩定性等困擾著人們的問題。本研究嚐試著提出一新的類神經網路學 習法則:解碼訓練法則 (Decode Algorithm), 並應用所研究推導出的訓練法則在字型辨認 的問題上。測試驗證的結果顯示,無論在訓練次數及字型辨認精確率的表現,較之現有法則 為優。 同時,依照所推導出來的解碼訓練法則所建立雙極性的加權值 (Weight),由於固定 且簡單,因此吾人深信其硬體 (hardware) 的實現性 (implementation) 是可預期的。 |
英文摘要 | There are a variety of types model to set up pattern and character recognition by applying the Artificial Neural Networks. However, among those networks, some shortcomings exist, such as the number of training iterations, parameters tuning, and its stability. In this work, we are concerned with the study of Decode Algorithm, a newly developed algorithm to the identification of characters. Results have shown that the Decode Algorithm works better than the other existing ones, especially in its training iterations as well as the exact rate in the area of character recognition. Meanwhile, we know that it is easy to decode the bipolar weights from the Decode Algorithm. Hence, we strongly believe that the possibility for the implementation of its hardware will be expected. |
本系統中英文摘要資訊取自各篇刊載內容。