Naive Bayes learners - Githubissues

vu-aml / adlib

Game-Theoretic Adversarial Machine Learning Library

MIT License

57 stars 16 forks source link

Naive Bayes learners #6

Closed jzyee closed 6 years ago

jzyee commented 6 years ago

I am trying to set up naive Bayes learners(bernoulli and multinomial), how do i set up the vectorizers for two separate situations?

presence or non presence in feature vector
the count of words stored in feature vector

I believe the load_dataset function is doing the 2nd situation currently by default, is this right?

jzyee commented 6 years ago

is it just the change in the binary parameter in the initialization of the EmailDataset?

xyvivian commented 6 years ago

EmailDataset() provides parameters to set binary and real value vectorizers. If binary flag is set to False, it is loading the presence of words stored in feature vector. The 2nd situation is by default when the binary flag is not set.
The robust naive bayes learner(Adversary Aware) has some scalability issues. We are fixing it.