How does Naïve Bayes behave under infinite feature limit asymptotics?

As feature dimensionality grows infinitely, Naïve Bayes performance depends on sparsity and independence validity.

In high-dimensional limits, Naïve Bayes can either improve or degrade depending on whether informative features remain sparse and independent. If noise dominates, likelihood products become unstable. However, in sparse regimes, informative signals dominate log-sum aggregation, stabilizing classification boundaries.