Sparse Matrix Compatibility and Optimized Vector Functions

[jknix]

Hey @joewandy, I've got a fork of this repo out for a project I'm working on, would you be open to me:

1. Adding python 3 support?
2. Optimizing some functions to use more efficient sparse arrays + vector operations in a few places? In my tinkering I'm getting a lot better performance

and then putting a PR back in?

[jknix]

Follow on, I'd also like to let the initial Hierarchical LDA object accept an initialized sklearn CountVectorizer object, that'd be a great compatibility.

[joewandy]

Sorry for the late reply (I'm travelling). I'd gladly accept any PR from you. Thank you so much for the contribution!!

[jknix]

Thanks @joewandy! I can't devote a ton of time, so don't expect anything immediate, but I'm making slow and steady progress.