國立高雄大學圖資館 |

登入

回首頁

Data-driven shrinkage strategies for quasi-regression.

紀錄類型:	書目-電子資源 : Monograph/item
正題名/作者:	Data-driven shrinkage strategies for quasi-regression.
作者:	Jiang, Tao.
面頁冊數:	90 p.
附註:	Adviser: Art B. Owen.
附註:	Source: Dissertation Abstracts International, Volume: 64-09, Section: B, page: 4444.
Contained By:	Dissertation Abstracts International64-09B.
標題:	Statistics.
電子資源:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3104254
ISBN:	0496518224

Data-driven shrinkage strategies for quasi-regression.
Jiang, Tao.

Data-driven shrinkage strategies for quasi-regression. [electronic resource] - 90 p.

Adviser: Art B. Owen.

Thesis (Ph.D.)--Stanford University, 2003.

In this thesis we aim to improve the performance of quasi-regression. To speed up the convergence of the estimates, we incorporate shrinkage parameters into procedures of coefficient estimation and function approximation. Two groups of shrinkage strategies are discussed and comparisons are made on a variety of examples. For black-box functions that have approximately sparse representations in the basis expansion, further enhancements are made with the adaptive basis search algorithm. It selects important basis functions into and removes unessential ones out of the model dynamically, making it possible to explore a even larger basis function set.

ISBN: 0496518224Subjects--Topical Terms:

182057
Statistics.

Data-driven shrinkage strategies for quasi-regression.
LDR:02519nmm _2200277 _450 001 162178
005 20051017073419.5
008 230606s2003 eng d
020 $a 0496518224
035 $a 00148679
035 $a 162178
040 $a UnM $c UnM
100 0 $a Jiang, Tao. $3 227295
245 1 0 $a Data-driven shrinkage strategies for quasi-regression. $h [electronic resource]
300 $a 90 p.
500 $a Adviser: Art B. Owen.
500 $a Source: Dissertation Abstracts International, Volume: 64-09, Section: B, page: 4444.
502 $a Thesis (Ph.D.)--Stanford University, 2003.
520 # $a In this thesis we aim to improve the performance of quasi-regression. To speed up the convergence of the estimates, we incorporate shrinkage parameters into procedures of coefficient estimation and function approximation. Two groups of shrinkage strategies are discussed and comparisons are made on a variety of examples. For black-box functions that have approximately sparse representations in the basis expansion, further enhancements are made with the adaptive basis search algorithm. It selects important basis functions into and removes unessential ones out of the model dynamically, making it possible to explore a even larger basis function set.
520 # $a Many statistical models and machine learning algorithms make use of black-box functions, i.e. complex functions depending on many variables in a poorly-understood fashion. Given such a function f(x) for d-dimensional x, it can be important to tell which variables, if any, dominate f and how they affect the function.
520 # $a The functional ANOVA decomposes a square integrable f into a sum of 2d component functions that depend on subsets of input variables. The variances of the component functions can be used to identify important variables and interactions. In quasi-regression the function f is expanded in an orthonormal basis. A large number of the coefficients are estimated by Monte Carlo simulation. The ANOVA component functions of f can be approximated as sums of coefficients times basis functions over certain indices sets. Numerical and graphical interpretation can be performed on the approximations.
590 $a School code: 0212.
650 # 0 $a Statistics. $3 182057
710 0 # $a Stanford University. $3 212607
773 0 # $g 64-09B. $t Dissertation Abstracts International
790 $a 0212
790 1 0 $a Owen, Art B., $e advisor
791 $a Ph.D.
792 $a 2003
856 4 0 $u http://libsw.nuk.edu.tw/login?url=http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3104254 $z http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=3104254