Supervised Machine Learning in Depth · Interactive activity

Naive Bayes

0 XP

A wrong assumption that works

It treats “free” and “prize” as unrelated words. That's obviously false. Yet it powered classic spam filters.

Why make such a “naive” assumption?