透過您的圖書館登入
IP:52.14.130.13
  • 學位論文

複雜資料應用流形學習之特徵值分析

Eigenvalues from Dimension Reduction on Data Complex

指導教授 : 劉長遠
若您是本文的作者,可授權文章由華藝線上圖書館中協助推廣。

摘要


近年來,因為資料量和資料維度的增加,降低資料維度成為重要的研究議題。若能將資料維度降低時亦可兼有保存資料原有特性的能力,對於資料的初步分析將有顯著的助益。在本論文中主要探討如何將流形學習的特徵值應用於資料分析。流形學習是降維演算法的一種分支,大部分此類的演算法都由三部分組成:首先需要解析出相鄰結構,接著利用相鄰結構計算出需要解的矩陣,最後經過特徵值分析將得到的特徵向量作為最後的降維結果輸出。藉由特徵值分析得到的特徵值,其實隱含了更多流形學習的細節。藉由觀察到特徵值的大幅度改變可以得到特定流形計算使用特定相鄰判定對於特定資料的有效參數範圍。而將不同的相鄰判定和矩陣運算方式代入計算同樣的資料,也會得到不同的降維結果。 本論文中將會以一些人工資料來驗證不同相鄰判定的可用性與特徵值變化對於最後降維結果的影響,也會將同樣的分析應用在股市、人臉、以及基因資料等真實的資料上。

並列摘要


Dimension reduction is an important issue since simple representation of dataset with large data dimension can be helpful for initial analysis. There are many variations of dimension reduction methods. In this work, the dimension reduction methods within a framework consists three parts are analyzed. The three parts of the frameworks are neighborhood relation construction, one matrix computed from the neighborhood relations, and the eigen-decomposition on the matrix in order to obtain eigenvectors as embedding results, while eigenvalues can be further analyzed to find out the effective range of parameters from different neighborhood selection approaches and different ways of matrix computation for obtaining reasonable embedding results. Since different neighborhood selection approaches can give different embedding results for different points of view on the specific dataset, the eigenvalue analysis for manifold learning methods within the framework can be useful for certain applications. In this work, some artificial datasets are for empirical analysis, while face dataset, stock dataset, and gene dataset will be used for true data analysis.

參考文獻


[1] Mikhail Belkin and Partha Niyogi. Laplacian eigenmaps for dimensionality
[2] Hong Chang and Dit-Yan Yeung. Robust locally linear embedding. Pattern
[3] Lisha Chen and Andreas Buja. Local multidimensional scaling for nonlinear
dimension reduction, graph drawing, and proximity analysis. Journal of the
American Statistical Association, 104(485):209–-219, 2009.

被引用紀錄


李丞倫(2014)。以支持向量機結合退火演算法及局部線性內嵌法模式推估降雨事件和河川警戒水位關係–以八掌溪為例〔碩士論文,國立臺灣大學〕。華藝線上圖書館。https://doi.org/10.6342/NTU.2014.00730

延伸閱讀