This fork is intended to extend lda to a form of disclda.
It amounts to a simple matrix manipulation of the topic vector for a document for that is based on the label of a document
In DIscLDA, each document has a label
Where in vanilla LDA you draw the word topics $z{d,i} ~ Cat(theta_d)$, in DiscLDA you draw $z{d,i} ~ Cat(T^{y_d}theta_d)$