一、雅可比(Jacobi)矩阵
对于函数
[y=f(x)
]
其中,(x=(x_1;x_2,...;x_n)),(y=(y_1;y_2;...;y_m))
则Jacobi矩阵为:
[ J=
egin{pmatrix}
frac{partial y_1}{partial x_1} & frac{partial y_1}{partial x_2} & frac{partial y_1}{partial x_3} & cdots & frac{partial y_1}{partial x_n} \
frac{partial y_2}{partial x_1} & frac{partial y_2}{partial x_2} & frac{partial y_2}{partial x_3} & cdots & frac{partial y_2}{partial x_n} \
vdots & vdots & vdots & ddots & vdots \
frac{partial y_m}{partial x_1} & frac{partial y_m}{partial x_2} & frac{partial y_m}{partial x_3} & cdots & frac{partial y_m}{partial x_n} \
end{pmatrix}
]
如果函数在一点(p)处可微,则Jacobi矩阵为函数在这一点处的最优线性逼近,即,
(f(x)approx f(p)+J(p)(x-p))
二、海塞(Hessan)矩阵
对于函数(f(x)),其中,(x=(x_1;x_2;x_3,...;x_n)),其Hessan 矩阵为:
[ H=
egin{pmatrix}
frac{partial f}{partial x_1partial x_1} & frac{partial f}{partial x_1partial x_2} & cdots & frac{partial f}{partial x_1partial x_n} \
frac{partial f}{partial x_2partial x_1} & frac{partial f}{partial x_2partial x_2} & cdots & frac{partial f}{partial x_2partial x_n} \
vdots & vdots & ddots & vdots \
frac{partial f}{partial x_npartial x_1} & frac{partial f}{partial x_npartial x_2} & cdots & frac{partial f}{partial x_npartial x_n} \
end{pmatrix}
]
Hessan matrix和Jacobi matrix关系:
[H_f=J(
abla f^T)
]
最优化中应用
当Hessan matrix正定时,在这一点取极小值;
当Hessan matrix负定时,在这一点取极大值;