http:^^www.cs.washington.edu^education^courses^590b^datafit^
来自「This data set contains WWW-pages collect」· EDU^EDUCATION^COURSES^590B^DATAFIT^ 代码 · 共 80 行
EDU^EDUCATION^COURSES^590B^DATAFIT^
80 行
Date: Mon, 02 Dec 1996 15:18:28 GMTServer: NCSA/1.4.2Content-type: text/html<html><title>Data Fitting Slides</title><body bgcolor="ffffff"><center><h1>Data Fitting</h1></center><hr noshade><blockquote><img src="t1.jpg"><blockquote><img src="t2.jpg"></blockquote><img src="t3.jpg"><blockquote><img src="t4.jpg"></blockquote><img src="t5.jpg"><blockquote><img src="t6.jpg"></blockquote><img src="t7.jpg"><blockquote><img src="t8.jpg"></blockquote><img src="t9.jpg"><blockquote><img src="t10.jpg"></blockquote><hr noshade>See Brad's slides here.<p><hr noshade><img src="t11.jpg"><blockquote><img src="t12.jpg"></blockquote>The a(i)'s have some probability distribution.Consider a(i) - atrue -- the real values are placed at the origin.This defines the uncertainties of a(0). <p>Goal: find the distribution of a(i) - atrue without knowingatrue or having infinite data sets.<p><img src="t13.jpg"><blockquote>Consider a world where a(0) is atrue. Hope that that world is not too different from the atrue world.In particular, assume that the probability distribution ofas(i)-a(0) is like a(i)-atrue.We can compute as(i)-a(0) and take that to be our answer.Each data set yields a point as(i)-a(0).Simulate enough sets to get a distribution of parameters.<p><img src="t14.jpg"></blockquote>You could analyze the cloud of points to get confidenceintervals for parameters independently.Or you could use a confidence region for all parameters simultaneously.You choose the level (68%) and shape.It is common to use constant chi-squared shapes around theD(0) minimum.<p>Point: you can learn something about the parameterconfidences by simulating the phenomenon.<p><img src="t15.jpg"><blockquote><img src="t16.jpg"></html>⌨️ 快捷键说明
复制代码Ctrl + C
搜索代码Ctrl + F
全屏模式F11
增大字号Ctrl + =
减小字号Ctrl + -
显示快捷键?