How to transform algorithms from Weka to Mahout?

[Weiwei Li <ha...@gmail.com>]

Hi Friends,
    I'm working on some algorithms in weka recently. As we know the benefits
of Mahout, I want to transform some algorithms from Weka to Mahout.
    Now I'm studying in C45.
    In Mahout, the classifier is only the Radom Forest Decision Tree, but
the method in it doesn't fit C45.
    So, I want to know weather should I write new Java code or not? If yes,
does it the same code as in Weka? Is there any document can help me?
    Thank you!

Re: How to transform algorithms from Weka to Mahout?

[Ted Dunning <te...@gmail.com>]

The goals of Weka and Mahout are very different.  Weka wants to have good
user interfaces and a very wide array of algorithms.  Mahout wants to have a
few high performance scalable algorithms.

So the first thing to look at is your own goals.  What do you need?

Secondly, is there something that you need to use Mahout for?  Is that
something that Mahout doesn't do?  I don't refer here to specific algorithm,
but rather to function.

Regarding your specific question, please do go ahead and write something for
Mahout.  I am not aware of any scalable implementation for C45, but it would
be interesting if you could do it.  This page may help you
https://cwiki.apache.org/confluence/display/MAHOUT/Developer+Resources

On Tue, Mar 1, 2011 at 9:18 AM, Weiwei Li <ha...@gmail.com> wrote:

> Hi Friends,
>    I'm working on some algorithms in weka recently. As we know the benefits
> of Mahout, I want to transform some algorithms from Weka to Mahout.
>    Now I'm studying in C45.
>    In Mahout, the classifier is only the Radom Forest Decision Tree, but
> the method in it doesn't fit C45.
>    So, I want to know weather should I write new Java code or not? If yes,
> does it the same code as in Weka? Is there any document can help me?
>    Thank you!
>