Hi Mark, Option B with Spark LDA is already underway, I am not sure of an ETA on the code changes.
- Nathanael > On Oct 28, 2016, at 2:00 PM, Mark Grover <[email protected]> wrote: > > Hi all, > Vartika mentioned this in a hallway conversation, we decided to bring it up > here. > > ONI's LDA implementation > <https://github.com/Open-Network-Insight/oni-lda-c/blob/01f1d6fa81ba0b3aa1080ca3a3ebd5c679720735/license.txt> > is > written in C, but most importantly, is GPL Licensed. And, from the looks of > it, it's a hard dependency. > > So, as we migrate to Apache, this needs to be changed. Either: > A. it needs to be a softer dependency, or > B. we need to find a substitute. > > I have noticed Spark MLLib has LDA, and given that this algorithm is being > called from Spark anyways, Option B (with Spark) would be a reasonable > choice. > > What do others think? Any other options? > > Mark
