where w
V = wHV w and ϕ
A = ϕTA−1ϕ. Similar to
[12], it can be proved that Algorithm 1 converges to optimum
if υ(t) converges.
Next, we need to find the condition that Lyapunov function
(23) converges. According to (16), ϕ(t + 1) is the projection
of ϕ(t)+AEq(t) onto ϕ 0. Hence the following inequality
holds (Prop. 3.2(b) in [19]),