Block-positivity of block-positive operator appears to yield incorrect result about qetlab HOT 2 CLOSED

Thanks for the report -- no user error on your part. That block positive matrix is right on the boundary of block positive matrices (i.e., if you minimize over product vectors, you get exactly zero), which makes it slightly tricky numerically. I had some numerical tolerance signs flipped. Fixed now. Now the following 3 lines of code yield results of -1, 1, and 0, respectively:

IsBlockPositive(Q0)
IsBlockPositive(Q0+0.001*eye(16))
IsBlockPositive(Q0-0.001*eye(16))

The first line yielding -1 (for "don't know") is a bit undesirable, but perhaps unavoidable: if the minimization over product states yields a value extremely close to 0, it's difficult to say with certainty whether or not it really is positive.

from qetlab.

@nathanieljohnston Thanks for the quick response! Glad to hear that there wasn't any user error on my part (this time, anyway :) ). And that explanation about the "don't know" case makes perfect sense. Thank you again for the quick response and fix!

from qetlab.