Gradient descent can be made online. But it's very slow and suffers from catastrophic forgetting. Typical gradient descent needs to iterate over the dataset man

Gradient descent can be made online. But it's very slow and suffers from catastrophic forgetting. Typical gradient descent needs to iterate over the dataset many times, while naive Bayes only needs one pass.