Re: [ojAlgo-user] Future of Jama

[Anders P. <ap...@op...>]

Can I ask you to join the ojAlgo mailing list?

https://lists.sourceforge.net/lists/listinfo/ojalgo-user

Dimitri Pourbaix wrote:
> Dear Anders,
>> Did you look at ojAlgo? What did you think?
> Yes, I did look at ojAlgo.  One thing that worries me (potentially a lot)
> is the overhead due to your 'Number' basic element rather than simply
> 'double'.  As I wrote in my very first mail, we are considering 6-month
> long intense computations.  So, any millisecond wasted in two much abs-
> traction can kill us.
Don't worry too much - look at the source code. Calculations are done 
using 'double' unless you choose to work with BigDecimal or 
ComplexNumber instead. I regularly benchmark ojAlgo against Jama - it 
performs very well!

http://ojalgo.org/generated/comparison.xml

Please write your own tests. If you want best possible performance and 
full control over which code is being executed you should work with 
MatrixStore instead of BasicMatrix.

http://ojalgo.org/generated/org/ojalgo/matrix/store/MatrixStore.html
http://ojalgo.org/generated/org/ojalgo/matrix/BasicMatrix.html

I recommend you get the latest source code from CVS. The only module you 
need is DevProj - it has no dependencies (other than Java5).

http://sourceforge.net/cvs/?group_id=79000
> The other aspect is that, if one restricts ojAlgo to the linear algebra
> part, I do not clearly see its advantage with respect to Jama.  The names
> of the methods are different ... but, in SVD for instance, I would not 
> get
> any functionality in its present version.  That is exactly where new 
> methods to the SVD part of ojAlgo would be very useful.  E.g., solving a
> overdetermined system of linear equation with SVD.
I've added a fair bit of "OO" (abstractions and extensibility) to ojAlgo 
compared to Jama, and I've manged to do that without loosing performance.

I also think that using BasicMatrix is as easy as using Jama, but you 
have the option to start using BasicArray and MatrixStore directly if 
need to.

The ojAlgo SVD interface gives you the pseudoinverse.

http://ojalgo.org/generated/org/ojalgo/matrix/decomposition/SingularValue.html#getPseudoinverse()
> If you want me to be more specific, let me know.
Give me code examples (typical to what you want to do) where Jama, or 
some other package, performs better and I'll try to "fix" the problem.

/Anders

> With my best regards,
>  Dim.