Mark,
Open source principles.
I agree with the principles, but IMHO, there are several issues with this release that are to fundamental to be released.
Furthermore, the different types of variances remains unsolved, as they are variant's of each other and should be in one single class.
Consensus was that these would be different classes because all Univariates are separate classes. PopulationVariance and SampleVariance should be two different classes in this case give the
prexising design of the package.
I disagree, because it are not two fundamental different statistics, just variants which do not deserve a seperate class. That would imply that any class in which sample as well as population variances can be used should be duplicated to deal with them, including classes as PCA's of which we then should produce in three variants, the third for correlations as a base for the calculations (gasp). But I know, it is the majority that want to keep it this way....
And I think the random class should be either eliminated (bad random) or be replaced with something more robust so that is actually usefull to be used for serious scientific work.
I would prefer to keep it and improve upon it in future versions of the library possibly integrating some RngPack and Colt classes. Again, we need a release out there before I can invest time into something like this.
What you essentially say with that is that people who want to use this package for serious scientif stuff should guess that the PNRG is very weak and should not be used.
Anyway, I encounter more and more programming issues with commons.math, but I do not report them anymore, because I do not want to go back to your source code.??? Thats inconsiderate, why are you commenting on if we should release if you have no interest in assisting in developing the package by reporting problems you encounter?
The problems I encounter are a direct result of some fundamental aspects of the package, such as different classes for variants of methods (variance, regression), which I treat as variants in my covariance matrix and PCA classes/methods. As those fundamantal aspects are apparently set, I need to work around them, which I do by making derived classes and packages that include aspects voted against here.
Well, if you don't contribute such changes back to the project as patches
Well, the reason for that is simple, my current source code for the Matrix.class is so fundamental different that making a patch would not reflect your version of it. The matrix class source contains 8 additional methods, and no implementation of the interface anymore. But if it is preferred that I do not provide feedback, that will be ok with me, but please let me know.
Please try to be flexible as well, this is what being part of a community is about, there is a "give and take" that needs to occur.
I tried, but the problem is as mentioned above, the source code where I make the changes in is to fundamental different to be used as a basis for patches.
Kim -- http://www.kimvdlinde.com
--------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
