I do not understand how the inequality in Example 2.1 in this paper is derived.
I do not understand the point in having:
$$-\lambda_{max}(\Sigma_u)$$
in the lower bound.
Using Weyl's inequality can we not get something better like: $$\lambda_{max}(\Sigma) > Var(f_t) b^Tb + \lambda_{min}(\Sigma_u)$$