Re: Convex Optimization / Quadratic Programming / Second Order Cone Programming

In fact, x' Qi x = 0 is a convex constraint, since this requires both
ways that Qi x = 0, which is a linear space (the nullspace of Qi.)
Ah oh, now I take back the both ways claim. The latter requires the
former, but not vice versa.. However, one can now say that, the
constraint is certainly convex if Qi is positive definite, having
enough to satisfy the "if and only if" condition.

What a confusion, sorry.

.

Relevant Pages