Ok, thanks. Let me take a look. That's interesting.

On Wed, Dec 29, 2010 at 3:49 PM, Ted Dunning <[email protected]> wrote:

> We have a hierarchy and as the clustering stuff begins to implement
> VectorClassifier, we will have more of it.
>
> The link function isn't pluggable right now, but that is a nit.
>
> On Wed, Dec 29, 2010 at 3:45 PM, Dmitriy Lyubimov <[email protected]>
> wrote:
>
> > I am not saying they are not doing that.  I am just saying there's a case
> > for generalizing Bayesian hierarchy instead of constraining it to two
> > particular models in a particular order with a particular link function.
> >
>

Reply via email to