- #1
divB
- 87
- 0
Hi,
Given an over-determined system of linear equations y=A c, the condition number of matrix A essentially says how good vector c can be restored from measurements y.
Changing the order of rows clearly does not change the condition number.
But is there information/literature on how to best choose a subset of rows to improve the condition number?
Say, e.g., vector c is small (30) but a large number of measurements are available (>>1000) and the goal is to reduce the number of measurements as much as possible (e.g. I only want to use 50 measurements). From experiments I find I get best results when just picking a random set of 50 rows. But is this really the case? Is there a way to proof this/understand this?
Given the fact I know the structure of the matrix A, is there a chance to find an optimal selection of rows?
Thanks
divB
Given an over-determined system of linear equations y=A c, the condition number of matrix A essentially says how good vector c can be restored from measurements y.
Changing the order of rows clearly does not change the condition number.
But is there information/literature on how to best choose a subset of rows to improve the condition number?
Say, e.g., vector c is small (30) but a large number of measurements are available (>>1000) and the goal is to reduce the number of measurements as much as possible (e.g. I only want to use 50 measurements). From experiments I find I get best results when just picking a random set of 50 rows. But is this really the case? Is there a way to proof this/understand this?
Given the fact I know the structure of the matrix A, is there a chance to find an optimal selection of rows?
Thanks
divB