These will be used as basic plumbing tests and therefore need to be implemented quickly. They should support the full set of functionality we expect (train, predict, calibrate) and will serve as templates for other extensions.

Note, some algorithms will support the concept of incremental training, instead of starting from scratch. We should make our data objects support this easily (i.e.: support an __add__ method) so we do not repeat large queries unnecessarily. Similarly, the parent classifier object may need to support an incremental_train method.

These need to be documented as part of the Sphinx pages for this project.