[Jboost-devel] A question about how to run alien (for example Weka) weak learners inside the JBoost

[Sina J. <si...@pr...>]

Dear all,

As a part of my research with Prof. Rob Schapire, I needed some 
experiments on combining boosting meta learning algorithms with some 
different base learners. As a result I decided to write a wrapper to 
connect JBoost to Weka. Right now, a class extending the Classifier 
class in Weka converts the data format from arff and runs a 
modification of the Connector class in JBoost, so I can run JBoost from 
weka right now (although it is still the primary version not dealing 
with many special cases). However, the challenging part is to run Weka 
classifiers as the weak learners in JBoost. I have written the 
converter to convert back from JBoost input and output formats to weka. 
However, I could not find out exactly how the special weak learning 
classifier (ADTree) works and what are the interfaces that have to be 
changed in order to be able to run alien (Weka) classifiers inside the 
JBoost code.
I will be very grateful if you help me with this.

Thanks in advance,
--Sina Jafarpour
http://www.cs.princeton.edu/~sina