have a section where we introduce the algorithms we use

[cnjr2]

i think this could be at the top of the document with an explanation why it makes sense to mix these models in particular.