What is the theoretical relationship between Naïve Bayes and maximum entropy models?

Naïve Bayes is a generative model, while maximum entropy models are discriminative with constraints on feature expectations.

Maximum entropy models choose the most uniform distribution satisfying feature constraints, whereas Naïve Bayes assumes a full generative structure of data. Both can yield similar linear decision boundaries under certain assumptions, but max-entropy models directly optimize conditional likelihood without modeling joint distribution.