矩阵看作线性算子进行诱导║A║ = max{║Ax║:║x║=1}= max{║Ax║/║x║:x≠0}