This code uses existing implementations of Naive Bayes (taken from Scikit Learn) to study the trend of generalization error with increasing number of examples.
Concretely two textual datasets are used:
- 20 newsgroups
- Reuters-21578
In particular the implementations of Naive Bayes analized are: Bernoulli and Multinomial
- Bernoulli Naive Bayes
- Multinomial Naive Bayes