查詢結果分析
來源資料
相關文獻
- A Kernel-based Fuzzy Compactness and Separation Clustering Algorithm
- 電化學雜訊於核能電廠蒸氣產生器化學清洗腐蝕現象之研究
- 以穩定性碳核素分析投餌施肥魚池中魚類生長能源的初步研究
- The Use of Stationary Hydroacoustic Transducer to Study Diel and Seasonal Influences on the Distribution of Fish School in Water Adjacent to the Cooling Water Intake of Nuclear Power Plant III
- 核醫藥物N,N'-1,2-ethanediylbis-cysteine diethylester dihydrochloride (ECD.2HCl)之合成
- Relationship Between the Dispersion of Thermal Effluent and the Tidal Current in the Waters Near the Outlet of the Third Nuclear Power Plant in Southern Taiwan
- 新能源政策推動現況與未來展望
- 穩健減核邁向低碳社會--新能源政策與減碳目標
- Taguchi-Method Assessment of Brain Perfusion Nuclear Medicine Scan of the Best Tuning Parameter
- 鎝環丁基丙烯胺肟非鏡像立體異構物合成路徑與影響因子之探討
頁籤選單縮合
題名 | A Kernel-based Fuzzy Compactness and Separation Clustering Algorithm=建構於緊緻和分隔的核模糊分類演算法 |
---|---|
作者姓名(中文) | 蔡旭琛; 徐翠琇; 周嘉榮; | 書刊名 | 德明學報 |
卷期 | 33:2 2009.12[民98.12] |
頁次 | 頁1-13 |
分類號 | 311.1 |
關鍵詞 | 模糊分類演算法; 模糊c分類; 模糊緊緻和分隔; 核; 穩健; 雜訊; Fuzzy clustering algorithms; Fuzzy c-means; FCM; Fuzzy compactness & separation; FCS; Kernel; Robust; Noise; |
語文 | 英文(English) |
中文摘要 | 聚類演算法主要是建構於群組內散佈矩陣使用緊緻測度。(例如 K-means, FCM等) 最近於 2005年吳國龍等提出一種新模糊分類演算法,稱為模糊緊緻和分隔演算法。此模糊緊緻和分隔演算法的目標函數是建構於模糊散佈矩陣,使得此演算法得自於同時考慮群內最小緊緻測度和群間最大分隔測度。2005年吳國龍等已經證明 FCS比 FCM更穩健於雜訊和異常值之干擾。雖然 FCS可以適度提升不受雜訊和異常值干擾之程度,但是它很依賴調整權重指數 m和參數。在我們實驗觀察下,我們發現FCS對於雜訊和異常值仍缺乏足夠穩健性,特別對於大小群之資料。在本論文,我們提出一種核型模糊緊緻和分隔分類演算法 (KFCS)。所提出的核型模糊緊緻和分隔分類演算法 (KFCS) 能改善 FCS之缺點,而且表現出更具效能和穩健性。數值和圖形例子之實驗用來展現 KFCS和 FCM及 FCS比較之效果。實驗結果顯示所提出KFCS 演算法比 FCM和 FCS演算法有更好之效果。 |
英文摘要 | Clustering algorithms are mainly based on a within-cluster scatter matrix with a compactness measure. (eg. K-means, FCM et al.). Recently, Wu et al. (2005) proposed a novel fuzzy clustering algorithm, called the fuzzy compactness & separation (FCS) algorithm. The FCS objective function is based on a fuzzy scatter matrix, such that the FCS algorithm can be derived by minimizing the compactness measure and simultaneously maximizing the separation measure. Wu et al. (2005) had also shown that FCS is more robust to noise and outliers than FCM. Although FCS can raise its insensitivity to noise and outliers to some extent, it somehow depends on the adjustment of the weighting exponent m and its parameters. On the basis of our experiments, we find that FCS still lacks enough robustness to noise and outliers, especially for unequal-sized-cluster data sets. In this paper, we propose a Kernel-based fuzzy compactness and separation clustering algorithm (KFCS). The proposed KFCS algorithm can improve the weakness in FCS and presents with more efficiency and robustness. Numerical and image experiments are performed to access the performance of KFCS in comparison with FCM and FCS. Experimental results show the proposed KFCS has better performance than FCS and FCM. |
本系統之摘要資訊系依該期刊論文摘要之資訊為主。