I'm writing a program for online versions of Bagging and AdaBoost algorithms and I'm using matlab's NaiveBayes classifier as the weak learner. Since as online learners they should get data one at a time, I have to update NaiveBayes classifier at each cycle, but it has no methods for it and I have to store all data in an array and call NaiveBayes.fit on hole array every time. Since I have about 4000 data and in real cases there may be many more I was wondering if you could tell me any better ways to do this? (Also if there are any other weak classifiers that I can look into please let me know)
Yours, Kourosh
You should run AdaBoost+NaiveBayes only on the new samples. The results can then be used to update the values of the previously found results.
You might also want to check out this version of a seemingly online version of NaiveBayes in MATLAB.