查詢結果分析
相關文獻
- 二維常態分配模型下之普適提研究
- 合歡山臺灣冷杉土壤之發育與分類
- 全民健康保險下疾病分類編碼品質與相關影響因素研究
- Fractal Image Coding Using Projection-Based Classification and Variable Shape Matching
- The Family Sapotaceae in Taiwan
- Genus Tisis Walker from Malaysia and Indonesia (Lepidoptera: Lecithoceridae), with Description of Three New Species
- Synopsis of the Genus Hygroplasta Meyrick in the World with Description of a New Species (Lepidoptera: Lecithoceridae)
- 談稅則分類的理論與實務
- 談稅則分類的理論與實務
- Ground Contact Pattern and Overuse Injury of the Lower Leg in Long Distance Runners
頁籤選單縮合
題名 | 二維常態分配模型下之普適提研究=Boosting Under Bivariate Normality |
---|---|
作者 | 吳尚勳; 曹振海; Wu, Shang-shiun; Tsao, Andy C.; |
期刊 | 中國統計學報 |
出版日期 | 20040900 |
卷期 | 42:3 2004.09[民93.09] |
頁次 | 頁259-275 |
分類號 | 319.3 |
語文 | chi |
關鍵詞 | 普適提; 訓練誤差; 一般誤差; 二維常態模型; 分類; Asset boosting; Bivariate normal models; Classification; Generalization error; Training error; |
中文摘要 | 普適提 (Boosting) 是近年來發展迅速且頗受歡迎的一種分類方法,但目前的理論結果尚無法完全解釋它在實際應用上的優良表現。在二維常態模型下,我們透過參數選擇來刻劃不同困難度的分類問題。在這個架構下以模擬實驗,我們研究普適提理論發展中的重要問題:weak base hypotheses的條件限制及其模型過過 ( overfitting) 的現象。我們發現:對於一般困難度的問題,普通提的確不太發生過過;另一方面,違反weak base hypotheses條件,在 一些情況下,間接導致普遍提的停止學習而使模型過適不致發生。 |
英文摘要 | Boosting is one of the most popular ensemble classifiers emerging in the past few years. Despite active researches, its excellent empirical performance is still much left unexplained. Under bivariate normal models, we characterize the classification problems of differential difficulty via parameter tuning. Under the settings, by simulation experimentation, we seek to gain understanding about the weak base hypotheses assumption and its implication on overfitting. The simulations suggest Boosting is resistant to overfitting in most cases. One enlightening observation. the violation of weak base hypotheses assumption, in some scenario, might lead the boosting into an idle state and in turn induces resistance to overfitting. |
本系統之摘要資訊系依該期刊論文摘要之資訊為主。