Re: DM: Applications of Canonical Correlations in Data Mining

[Anthony Rossini]

>>>>> "Krishnadas" == Krishnadas  <ckkrish@cyberspace.org> writes:


    Krishnadas> I would like to know if there have been any
    Krishnadas> application of canonical correlations in datamining
    Krishnadas> problems. Would appreciate pointers to
    Krishnadas> papers/references, software, success stories.

I'd be a bit worried about the underlying assumptions which were used 
to develop CC.  

On the other hand, there have been some interesting semi/infinite
parametric methods developed in recent years to solve similar 
problems 
(principle(?) curves, for principal(?) components, and other smoothed
versions of classical multivariate statistical methods).  Assuming
that the computational complexity is solvable (hard assumption!  but
bear with me), has anyone considered these methods?
Papers/references/thoughts appreciated.

Similarly, has anyone worked with "statistical optimality" in the
context of data mining?  (ala Bickel,Klassen,Ritov, and Wellner,if 
you 
are familiar with the book); I'm thinking in terms of comparing
methods in terms of efficiency of information extraction (efficient 
in 
terms of  "statistical amount", not computational issues).

best,
-tony