Derive the normalized LMS whose objective is to minimize the change of the weight vector
$||\delta w(n+1)||$ subject to the constraint that $u^H(n)w(n+1) = d(n)$. Here, the filter input is
$u^H(n) = [u(n),u(n-1),……,u(n-M+1)]$ and $u(n)$ represents the observation at time $n$.