AdaptiveLogisticRegression includes the following note about
beingworthwhile for binary values only. Is this still true, or are
therealternative crossfold learners for multinomial data?
org.apache.mahout.classifier.sgd.AdaptiveLogisticRegression----------
* The fitness used here is AUC.  Another alternative would be to try
log-likelihood, but it is much * easier to get bogus values of
log-likelihood than with AUC and the results seem to accord pretty *
well.  It would be nice to allow the fitness function to be pluggable.
This use of AUC means that * AdaptiveLogisticRegression is mostly
suited for binary target variables. This will be fixed * before long
by extending OnlineAuc to handle non-binary cases or by using a
different fitness * value in non-binary cases.---------------

-- 
Lance Norskog
[email protected]

Reply via email to