How does Naïve Bayes perform feature selection implicitly in high-dimensional spaces?

Naïve Bayes implicitly downweights irrelevant features through probability multiplication and smoothing.

Features with low class-conditional likelihood contribute minimal influence to posterior scores. Over time, irrelevant features get naturally suppressed because they do not significantly change P(X|C). This acts as an implicit feature selection mechanism, especially in text classification where noisy words exist.