A practical question about classification

Hello,


Let's say we have the output of a hierarchical clustering algorithm on
the attributes of trainining pairs, except class of course, we will
use for classification. How would you use a neural network to turn
this information to a classifier, or boost an existing classifier?

A bit of explanation because it may seem pointless at first. The
classification might be a non-linear function of a "natural"
categorization selected according to a priori, unrelated criteria.
Still, I suppose we can expect to extract information from this
mapping. What do you think?

Thanks,

--
Eray Ozkural