Re: uBLAS - Matrix Inversion

From: speedsnaii speedsnaii_at_[hidden])
Date: 2003-02-25 09:16:44

> Hello Thomas,
> A couple of months ago I had the exact same query.  Unfortunatly, 
> uBLAS does not perform matrix inversion.  There are essentially two 
> options for you.  #1: I have found a couple of matrix libraries 
that 
> are not of the same level of quality as uBLAS but suit ones needs 
to 
> a reasonable degree.  You can find these through google, just type 
> in C++ matrix or something to that effect and you will find many of 
> them.  #2: Obtain "Numerical Recipies in C++."  This book is the 
C++ 
> version of the very popular Numerical Recipies in C book, and 
> details a lot of frequently used scientific algorithms.  You would 
> then of course have to implement these ideas.
> 

I am just in the course of learning UBLAS. But I've seen the 'solve' 
functions. I expect it should be possible to do matrix inversion with 
these. Matrix inversion is nothing more than solving for a matrix 
of 'right hand' vectors. Normally you don't really need to invert a 
matrix at a whole since for practical purposes it is much faster to 
save the decomposed (L-R) parts and then backsupstitute (solve) for 
each right hand vector. So I think this is reflected in UBLAS.

But I also would like to see a small example program of how to do 
this with UBLAS.

Good luck,
Roland