Currently SpamBrainz only supports offline learning which requires that models be retrained on the entire dataset.

Ideally it should be possible to continue training an existing model with new data while still maintaining high training efficiency.