Re: [Bayes++] Usage clarification for linear algebra classes

[Michael S. <ma...@mi...>]

On Wednesday 09 May 2007 17:36, yon wrote:
> Dear Michael, Dear List,
>
> Can someone provide a quick indication of usage compatibility and
> conversion between the different linear algebra classes?
>
> Specifically, which of FM::Vec, ublas::vector<float>, and the bayes++
> column_matrix implementations are compatible, and how to convert
> between them?  I became stuck on this and re-implemented two
> functions for which I could not figure out the conversions, but it
> would be better that I handle the conversions correctly.


'FM::Vec' is derived from 'ublas::vector<double>'.
It is therefore type compatible with a ublas::vector<double> but NOT with
a ublas::vector<float>.

I assume by 'column_matrix' you mean a 'ublas::matrix_column'? 
The uBlas 'column( matrix, col )' function allows you to create 
a 'matrix_column' which references a given matrix. A good reference for the 
type of usage possible look at SIRFlt.cpp itself.

The 'matrix_column' class has no type relationship with any vector class. 
However the generic assignment and expression operations in uBLAS allow you to 
use it where a the other vector classes can be used.

You own code will also have to be generic (use templates) if it want to be 
able to operate on the different types as defined by the uBLAS Vector 
concept.

Michael
-- 
___________________________________
Michael Stevens Systems Engineering

34128 Kassel, Germany
Phone/Fax: +49 561 5218038

Navigation Systems, Estimation  and
                 Bayesian Filtering
    http://bayesclasses.sf.net
___________________________________