Any who have followed this project know that LDA can be a bit of a black-art to implement. Any algorithm base on probabilities shares this "how do you validate it" problem. Today, though, the results where so dramatically more inline with what the research papers were claiming that I'm fairly sure we'll have some solid working implementation out soon.

This release will break K-Mean clustering in that the APIs will differ very slightly, but enough to break one of the engine scripts. However, it should be an afternoon of touch-up work to bring K-Mean back in line with LDA's API changes.

The LDA call that will cause the problem can just be commented out of the engine script to use K-Mean.

Latest and greatest is, of course, in the repository.