Maiej,

With regard to semi-supervised learning, we have taken baby steps in getting
all of our classifiers and clustering algorithms to speak a common language.

The goal is to have a classifier be usable in all contexts, regardless of
whether it came from supervised, semi-supervised or unsupervised origins.

That is an area where work is needed.

We have no reinforcement learning or active learning capabilities.  Those
would also be areas of some potential, especially the active learning.
 Vowpal Wabbit just added an active learning component which is reputedly
doing very well.


2010/12/27 Maciej Jaskowski <[email protected]>

> Thank you for all answers!
>
> @Shannon: thank you for indicating those. I have skimmed them and I
> feel that it would be much better if someone encouraged me to
> undertake a particular problem on the basis of my interests.
> I mean, the list you mentioned is much more meaningful to you, then to me
> ;)
>
> @Dmitriy: it might disappoint you but I have none :/ All I have is my
> own Mac with Hadoop 0.20.2 installed.
>
> @Ted: My interests are in supervised and semi-supervised learning in
> particular right now.
> Reinforcement Learning and Recommendations are stuff which I would
> like to learn, so if there is some stuff to do there, I will take it
> willingly!
>
> Greetings,
> Maciej.
>
> On 25 December 2010 20:59, Ted Dunning <[email protected]> wrote:
> > Maciej,
> >
> > Can you say what area of interest you have?  Mahout has recommendation,
> > clustering, classification, large data dimensionality reduction and other
> > capabilities.
> >
> > Does one strike you as particularly interesting yet?
> >
> > On Sat, Dec 25, 2010 at 2:05 AM, Maciej Jaskowski <
> > [email protected]> wrote:
> >
> >> Hi everyone,
> >>
> >> I will have some time (say half of each week) from mid January to mid
> >> March for a hard work on some interesting OS project.
> >>
> >> Since I have a fine 3-months experience in developing using Hadoop and
> >> I am interested in Machine Learning, Mahout seems to me like an
> >> excellent choice :-)
> >>
> >> So... how can I help?
> >>
> >> Greetings,
> >> Maciej Jaśkowski
> >>
> >
>

Reply via email to