[mlpack-svn] [MLPACK] #344: Using Welford method to calculate variance in Naive Bayes Classifier

[MLPACK Trac]

#344: Using Welford method to calculate variance in Naive Bayes Classifier
-------------------------+--------------------------------------------------
 Reporter:  akvah        |        Owner:                      
     Type:  enhancement  |       Status:  new                 
 Priority:  minor        |    Milestone:                      
Component:  mlpack       |     Keywords:  variance calculation
 Blocking:               |   Blocked By:                      
-------------------------+--------------------------------------------------
 Hi,

 I have created a little patch to calculate the variance in the naive bayes
 classifier using the Welford method. This method is more robust for
 variance calculation when the number or amount of values are large. e.g.
 in the previous implementation, the variance between
 100000000,100000001,100000002 would be zero.
 For more information look at:
 http://en.wikipedia.org/wiki/Algorithms_for_calculating_variance#Incremental_Algorithm

 Vahab

-- 
Ticket URL: <https://trac.research.cc.gatech.edu/fastlab/ticket/344>
MLPACK <www.fast-lab.org>
MLPACK is an intuitive, fast, and scalable C++ machine learning library developed at Georgia Tech.