线性矩阵不等式的常用引理：
Lemma 1： M M M 是对称阵，那么
λ max ( M ) ≤ t ⟺ M − t I ≤ 0 \lambda_{\max }(M) \leq t \Longleftrightarrow M-t I \leq 0 λmax(M)≤t⟺M−tI≤0
Proof. Note that for an arbitrary matrix M M M with an eigenvalue s s s and a corresponding eigenvector x , x, x, there holds
( M − t I ) x = M x − t x = ( s − t ) x (M-t I) x=M x-t x=(s-t) x (M−tI)x=Mx−tx=(s−t)x
This states that for an arbitrary matrix M M M there holds
λ ( M − t I ) = λ ( M ) − t \lambda(M-t I)=\lambda(M)-t λ(M−tI)=λ(M)−t
Thus, when M M M is symmetric, we have
λ max ( M ) ≤ t ⟺ λ max ( M − t I ) ≤ 0 ⟺ M − t I ≤ 0 \begin{aligned} \lambda_{\max }(M) \leq t & \Longleftrightarrow \lambda \max (M-t I) \leq 0 \\ & \Longleftrightarrow M-t I \leq 0 \end{aligned} λmax(M)≤t⟺λmax(M−tI)≤0⟺M−tI≤0
Lemma 2： A A A 是具有合适维数的矩阵， t t t 是一个正数，那么
A T A − t 2 I ≤ 0 ⟺ [ − t I A A T − t I ] ≤ 0. A^{\mathrm{T}} A-t^{2} I \leq 0 \Longleftrightarrow\left[\begin{array}{cc}-t I & A \\ A^{\mathrm{T}} & -t I\end{array}\right] \leq 0. ATA−t2I≤0⟺[−tIATA−tI]≤0.
Proof: Put
Q = [ I A 0 t I ] Q=\left[\begin{array}{ll} I & A \\ 0 & t I \end{array}\right] Q=[I0AtI]
then Q Q Q is nonsingular since t > 0. t>0 . t>0. Note that
Q T [ − t I A A T − t I ] Q = [ − t I 0 0 t ( A T A − t 2 I ) ] Q^{\mathrm{T}}\left[\begin{array}{cc} -t I & A \\ A^{\mathrm{T}} & -t I \end{array}\right] Q=\left[\begin{array}{cc} -t I & 0 \\ 0 & t\left(A^{\mathrm{T}} A-t^{2} I\right) \end{array}\right] QT[−tIATA−tI]Q=[−tI00t(ATA−t2I)]
again in view of the fact that t > 0 , t>0, t>0, it is clearly observed from the above relation that the equivalence in ( 1.2.4 ) (1.2 .4) (1.2.4) holds. Based on the above lemma, we can derive the following conclusion.
参考资料：
3066
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
