Question on the discrete part v and G

[drjingma]

Hi Andrew, I learned your work recently and am curious as to whether the method still works if some genes are "on" all the time. Using your notations in the paper, this amounts to say that there exists some i, where i stands for variable index, the random variable v_i=1 with prob 1. (This may not be true in scRNA, but let's say hypothetically.)

If there's only the discrete part, the above case may lead to problems estimating the matrix G (the Ising model part). Did you have to enforce certain identifiability condition in your work?

Thanks!