Model Training & Adaptation
Naive Bayes
A group-prediction method that combines clues while assuming they act independently within each group.
Example
A text prediction system combines word evidence to identify spam.
Why people use it
It offers a relatively simple way to combine clues when predicting a category.
What you'll hear
“Use the words as clues about the message's category.”
What this means for you
Compare actual performance rather than accepting or rejecting the method on its name alone.
Can you control it?
Developer-only
The people building or running the AI choose this setup. An everyday user generally needs their help to change how this part works.
Common questions
- Can it work even when its independence assumption is imperfect?
- Yes. It can still give useful predictions, although closely related clues can distort its estimated chances.
- Can it be useful with limited computing resources?
- Yes. Its relatively simple calculations can make it quick to train and use.
- Can an unseen word affect a text classifier?
- Yes. The system needs rules for handling clues that did not appear in its teaching examples.