Can you put your code on github? There is a detail that slipped somewhere and I can't guess where it is. Your constructor is correct for a binary classifier, but I can't say much else.
How much data, btw, did you pour in? On Sun, May 1, 2011 at 3:18 AM, Tim Snyder <[email protected]> wrote: > I am currently using trunk from April 30, 2011. The code is loosely > following the SGD training example from Mahout in Action. I have > instantiated the learner with the purpose of having a binary classifier > : > > AdaptiveLogisticRegression learningAlgorithm = new > AdaptiveLogisticRegression( 2, FEATURES, new L1() ); > > Everything works fine (ie. the training) until I get to > learningAlogrithm.close() where I get the following exception: > > learningAlgorithm.close() Error = > java.lang.ArrayIndexOutOfBoundsException > java.lang.IllegalStateException: > java.lang.ArrayIndexOutOfBoundsException > Exception = null > at > > org.apache.mahout.classifier.sgd.AdaptiveLogisticRegression.trainWithBufferedExamples(AdaptiveLogisticRegression.java:144) > at > > org.apache.mahout.classifier.sgd.AdaptiveLogisticRegression.close(AdaptiveLogisticRegression.java:196) > at > > com.zensa.spinn3r.mahout.SnomedSDGClassificationTrainer.finishTraining(SnomedSDGClassificationTrainer.java:159) > at > > com.spinn3r.sdg.trainer.Spinn3rSDGTrainer.process(Spinn3rSDGTrainer.java:170) > at > com.spinn3r.sdg.trainer.Spinn3rSDGTrainer.main(Spinn3rSDGTrainer.java:272) > Caused by: java.lang.ArrayIndexOutOfBoundsException > > If I change the number of categories to 100, the close() works fine. Any > ideas on how to get around this and have a working binary classifier? > > Thanks in advance. > > Tim > >
