Naive Bayes parameter question

kathi546

Hi everybody!
I´m working with RapidMiner and don´t understand the naiveBayes-parameter "use_weights" and "use_kernels". What is the different between them? I read the tutorial but my confusion remains. Please, could anybody be so kind to help me?
many thanks and best wishes

kathi

PS: I already posted that question in the general community forum, but I think here is the better place to ask. sorry for any inconvenience. :-\

land

Hi Kathi,
the "use_weights" parameter enables the learner to use example weigths if provided. Example weights contain information about the importance of every single example. This might be used to prevent very important (and probably costly) examples to be missclassified.
This parameter is available in several learning algorithms.
"use_kernels" has in fact been removed in version 4.3, since NaiveBayes then becomes something of a k nearest Neighbour learner, needing to store the complete training set. But exactly this behavior is prohibited in areas the original naive bayes is suited best for: Large Datasets.
If you use an old version, the parameter referes to Flexible Bayes and implements the idea of http://staff.icar.cnr.it/manco/Teaching/2006/datamining/articoli/flex.uai95.pdf.

Greetings,
  Sebastian