Robin, (and everybody) Watch out for pending wide-scale changes from MAHOUT-480.
Anybody who would like to avoid those changes hanging fire can do a quick (right) review of the pending patch. Only 93 files are changed so it should be a breeze! Eugen did a ton of work to get all those changes done in the first place. I found a few discrepancies and would like somebody to check my checking. I will commit the changes as soon as I am sure we got these right. On Sat, Oct 2, 2010 at 3:35 PM, Robin Anil <[email protected]> wrote: > I just uploaded a patch for moving DictionaryVectorizer. Take a look > > On Sun, Oct 3, 2010 at 1:09 AM, Robin Anil <[email protected]> wrote: > > > Ok. > > > > > > On Sun, Oct 3, 2010 at 12:58 AM, Ted Dunning <[email protected] > >wrote: > > > >> Or just flat in o.a.m.vectorizer. There aren't a lot of classes we are > >> talking about. > >> > >> On Sat, Oct 2, 2010 at 12:26 PM, Robin Anil <[email protected]> > wrote: > >> > >> > i.e > >> > o.a.m.vectorizer.dictionary > >> > o.a.m.vectorizer.hashencoder > >> > > >> > > >> > On Sun, Oct 3, 2010 at 12:49 AM, Ted Dunning <[email protected]> > >> > wrote: > >> > > >> > > Same question applies to the FeatureValueEncoder family currently > >> under > >> > > classifier.vectors.* > >> > > > >> > > On Sat, Oct 2, 2010 at 11:23 AM, Robin Anil <[email protected]> > >> > wrote: > >> > > > >> > > > How do you feel like moving the DictionaryVectorizer and Colloc > >> > generator > >> > > > to > >> > > > the Core under vectorizers package instead of keeping them under > >> utils. > >> > > > FeatureEncoders will also be moved under vectorizers. I want to > add > >> a > >> > > > Wrapper which takes a Vectorizer and converts input data to > vectors. > >> > Its > >> > > > the > >> > > > missing piece of the Classifier puzzle > >> > > > > >> > > > o.a.m.vectorizer.dictionary > >> > > > o.a.m.vectorizer.hashed or something funkier? > >> > > > > >> > > > What do you think about this? > >> > > > > >> > > > Robin > >> > > > > >> > > > >> > > >> > > > > >
