附有奇異值修正限制的改進的嶺估計方法
發(fā)布時間:2018-08-25 19:21
【摘要】:最小二乘估計具有無偏性,而嶺估計是一種有偏估計,它通過引入偏差降低方差來降低均方誤差。在模型出現(xiàn)病態(tài)時,嶺估計優(yōu)于最小二乘估計。對嶺估計的方差與偏差進行分析發(fā)現(xiàn),嶺估計通過修正病態(tài)矩陣的奇異值降低均方誤差,但對部分較大奇異值的修正不能有效降低均方誤差。通過比較修正奇異值的方差下降量與偏差引入量的大小關系確定需要修正的小奇異值,進而改進嶺估計方法,實現(xiàn)選擇性地修正小奇異值,提出附有奇異值修正限制的改進的嶺估計方法,可有效改善嶺估計的解算效果和可靠性,實驗驗證了新方法的可行性和有效性。
[Abstract]:The least square estimation is unbiased, while the ridge estimation is a kind of biased estimation, which reduces the mean square error by introducing the deviation to reduce the variance. When the model is ill-conditioned, the ridge estimation is superior to the least square estimation. By analyzing the variance and deviation of ridge estimation, it is found that ridge estimation reduces mean square error by modifying singular value of ill-conditioned matrix, but correction of partial larger singular value can not effectively reduce mean square error. By comparing the relationship between the variance drop of the modified singular value and the deviation introduced quantity, the small singular value that needs to be modified is determined, and then the ridge estimation method is improved to realize the selective correction of the small singular value. An improved ridge estimation method with singular value correction constraints is proposed, which can effectively improve the solution effect and reliability of the ridge estimation. The experimental results show that the new method is feasible and effective.
【作者單位】: 中南大學地球科學與信息物理學院;
【基金】:國家自然科學基金(41531068,41474008) 國家重點基礎研究發(fā)展計劃(2013CB733303)~~
【分類號】:P207
本文編號:2203857
[Abstract]:The least square estimation is unbiased, while the ridge estimation is a kind of biased estimation, which reduces the mean square error by introducing the deviation to reduce the variance. When the model is ill-conditioned, the ridge estimation is superior to the least square estimation. By analyzing the variance and deviation of ridge estimation, it is found that ridge estimation reduces mean square error by modifying singular value of ill-conditioned matrix, but correction of partial larger singular value can not effectively reduce mean square error. By comparing the relationship between the variance drop of the modified singular value and the deviation introduced quantity, the small singular value that needs to be modified is determined, and then the ridge estimation method is improved to realize the selective correction of the small singular value. An improved ridge estimation method with singular value correction constraints is proposed, which can effectively improve the solution effect and reliability of the ridge estimation. The experimental results show that the new method is feasible and effective.
【作者單位】: 中南大學地球科學與信息物理學院;
【基金】:國家自然科學基金(41531068,41474008) 國家重點基礎研究發(fā)展計劃(2013CB733303)~~
【分類號】:P207
【相似文獻】
相關期刊論文 前10條
1 李屹旭;張俊;;奇異改進型嶺估計及其在大地測量中的應用[J];貴州大學學報(自然科學版);2007年02期
2 黃幼才;嶺估計及其應用[J];武漢測繪科技大學學報;1987年04期
3 汪曉慶;張方仁;;廣義嶺估計及其在測量平差中的應用[J];武測科技;1991年01期
4 隋立芬;抗差嶺估計原理及其應用[J];測繪通報;1994年01期
5 歸慶明,黃維彬,張建軍;抗差泛嶺估計[J];測繪學報;1998年03期
6 劉雁雨;李建偉;劉曉剛;;部分嶺估計的嶺參數(shù)確定方法[J];測繪科學技術學報;2007年06期
7 鄭進鳳,郭宗河;研究和應用嶺估計時值得注意的問題[J];測繪工程;1995年01期
8 馮光財;戴吾蛟;朱建軍;陳正陽;;基于虛擬觀測的病態(tài)問題解法[J];測繪科學;2007年02期
9 隋立芬;抗差嶺估計的誤差影響測度[J];測繪學報;1995年02期
10 王彬;高井祥;劉超;王堅;周鋒;;L曲線法在等價權抗差嶺估計模型中的應用[J];大地測量與地球動力學;2012年03期
,本文編號:2203857
本文鏈接:http://sikaile.net/kejilunwen/dizhicehuilunwen/2203857.html
最近更新
教材專著