Mathematica 9 is now available
Student Support Forum
-----
Student Support Forum: 'Regarding least norm solutions' topicStudent Support Forum > General > Archives > "Regarding least norm solutions"

Help | Reply To Topic
Author Comment/Response
Patrik
10/02/11 1:13pm

Hello. I am posed a linear optimazation problem in the form of Ax = y, where I need to find the best solution x for the matrix-equation.

The properties of the matrix A is that it's fat (more coulmns than rows), also there is no guarantee that the columnspace of A is linearly independent(that is, A may not have full rank).

Trouble is, when A doesnt not have full rank, there is a set of non-uniqe solutions x and I need to pick the best one out of these. This is my problem.

If A does have full rank I can just

A := PseudoInverse[~some matrix elements];
y := {~some matrix elements};
A.y

and evaluate (this is the least norm solution). Then I am guaranteed the least norm solution.

So, is there some extension to this so I can handle it when A does not nessecarily have full rank? I was told Matlab has this, tho I'm using Mathematica, so here I am.

URL: ,
Help | Reply To Topic