Bregman Divergences (Mahalanobis Distance) in Cross Distances operator not working

amitd

The Bregman Divergences (Mahalanobis Distance) in the Cross Distances operator. Given below is the stack trace.

Exception: java.lang.NullPointerException

Message: null

Stack trace:

  Jama.Matrix.times(Matrix.java:742)

  com.rapidminer.tools.math.similarity.divergences.MahalanobisDistance.calculateDistance(MahalanobisDistance.java:49)

  com.rapidminer.tools.math.similarity.BregmanDivergence.calculateSimilarity(BregmanDivergence.java:54)

  com.rapidminer.tools.math.similarity.DistanceMeasure.calculateSimilarity(DistanceMeasure.java:267)

  com.rapidminer.operator.similarity.CrossDistancesOperator.doWork(CrossDistancesOperator.java:156)

  com.rapidminer.operator.Operator.execute(Operator.java:1023)

  com.rapidminer.operator.execution.SimpleUnitExecutor.execute(SimpleUnitExecutor.java:77)

  com.rapidminer.operator.ExecutionUnit$2.run(ExecutionUnit.java:805)

  com.rapidminer.operator.ExecutionUnit$2.run(ExecutionUnit.java:800)

  java.security.AccessController.doPrivileged(Native Method)

  com.rapidminer.operator.ExecutionUnit.execute(ExecutionUnit.java:800)

  com.rapidminer.operator.OperatorChain.doWork(OperatorChain.java:423)

  com.rapidminer.operator.Operator.execute(Operator.java:1023)

  com.rapidminer.Process.executeRoot(Process.java:1464)

  com.rapidminer.Process.lambda$executeRootInPool$5(Process.java:1443)

  com.rapidminer.studio.concurrency.internal.AbstractConcurrencyContext$AdaptedCallable.exec(AbstractConcurrencyContext.java:362)

  java.util.concurrent.ForkJoinTask.doExec(ForkJoinTask.java:289)

  java.util.concurrent.ForkJoinPool$WorkQueue.runTask(ForkJoinPool.java:1056)

  java.util.concurrent.ForkJoinPool.runWorker(ForkJoinPool.java:1692)

  java.util.concurrent.ForkJoinWorkerThread.run(ForkJoinWorkerThread.java:175)

MartinLiebig

Hi,

can you please provide an example process. this needs to be something specific, since MahalnobisDistance is used quite a lot. My thought is an uninvertable CovarianceMatrix, which results in a NPE rather than a readable usererror.

BR,

Martin

amitd

I have attached a sample process with Iris dataset.