透過您的圖書館登入
IP:18.188.146.77
  • 學位論文

Sub-Gaussian 隨機矩陣的大偏差理論

Deviation Inequalities of Sub-Gaussian Random Matrix

指導教授 : 許元春

摘要


隨機矩陣在資料科學中扮演重要的角色,其一是隨機矩陣提供了一個降維的 方法,因為在應用上我們觀測到的資料維度過高,使我們不易分析也不好觀察資料的結構,我們的解決方法是將隨機矩陣視為隨機映射,其可以將高維度資料投射至低維度空間,且有很高的機率降維後的資料可以維持其原本的結構,為了說明這個現象,我們需要研究有關隨機矩陣的大偏差估計。

並列摘要


We devote to an exploration of sub-gaussian random matrices (i.e. a random matrix with i.i.d mean-zero, unit variance, sub-gaussian entries) in a non-asymptotic setting, with the goal of obtaining explicit deviation inequalities. Therefore, we can use these deviation inequalities to show that sub-gaussian random matrix can be regarded as a random projection, which project from high-dimensional data to lowdimension space and preserve their relative distance.

參考文獻


[1] Gideon Schechtman, Two observations regarding embedding subsets of Euclidean
spaces in normed spaces. Advances in Mathematics, Volume 200, Issue
1, 15 February 2006, Pages 125-135.
[2] Roman Vershynin, High-Dimensional Probability: An Introduction with Applications
in Data Science. Cambridge University Press, 2018.

延伸閱讀