How does Naïve Bayes behave under heteroscedastic feature distributions?

Naïve Bayes performance degrades under heteroscedasticity when feature variance differs significantly across classes.

Heteroscedasticity refers to varying feature variance across classes. Gaussian Naïve Bayes assumes class-conditional distributions with specific means and variances. When variance differs drastically or is misestimated, likelihood ratios become unstable, distorting decision boundaries and posterior probabilities.