从Schrödinger方程和Heisenberg方程的等价性看Kalman滤波器和Gauss-Newton的等价性

Schrödinger equation:
$-i\hslash \frac{\partial \psi }{\partial t}=\hat{H}\psi$
Heisenberg equation:
$\dot{a}(t)=Ha(t)$

from Schrödinger to Heisenberg:
Let $\psi ={\Sigma }_{i=0}^N{a}_i(t){\varphi }_i(\vec{r})$
then: $\frac{\partial \psi }{\partial t}={\Sigma }_{i=0}^N{\dot{a}}_i(t){\varphi }_i(\vec{r})$
then: $<\frac{\partial \psi }{\partial t},{\varphi }_k^{\ast }>={\dot{a}}_k(t)$
then: $<\hat{H}\psi ,{\varphi }_k^{\ast }>={\Sigma }_{i=0}^N{a}_i(t)<\hat{H}{\varphi }_i,{\varphi }_k^{\ast }>$
in all:
$$\dot{\left[\begin{array}{c}{a}_0\\ a_1\\ ...\\ a_N\end{array}\right]}=\left[\begin{array}{c}<\hat{H}{\varphi }_0,{\varphi }_0^{\ast }>,<\hat{H}{\varphi }_1,{\varphi }_0^{\ast }>,<\hat{H}{\varphi }_2,{\varphi }_0^{\ast }>,...,<\hat{H}{\varphi }_N,{\varphi }_0^{\ast }>\\ <\hat{H}{\varphi }_0,{\varphi }_1^{\ast }>,<\hat{H}{\varphi }_1,{\varphi }_1^{\ast }>,<\hat{H}{\varphi }_2,{\varphi }_1^{\ast }>,...,<\hat{H}{\varphi }_N,{\varphi }_1^{\ast }>\\ ...\\ <\hat{H}{\varphi }_0,{\varphi }_N^{\ast }>,<\hat{H}{\varphi }_1,{\varphi }_N^{\ast }>,<\hat{H}{\varphi }_2,{\varphi }_N^{\ast }>,...,<\hat{H}{\varphi }_N,{\varphi }_N^{\ast }>\end{array}\right]\left[\begin{array}{c}{a}_0\\ a_1\\ ...\\ a_N\end{array}\right]$$
Kalman Filter:
$X_{k+1}=A{X}_k+B{u}_k+w$
$Y_k=H{X}_k+v$
$w$~N(0,${\sigma }_1$)
$v$~N(0,${\sigma }_2$)

Gauss-Newton:
1-dimensional:
y=f(x)
$$\begin{gathered} y_1-y_0=f(x_1)-f(x_0) \\ =f^{{}^{\prime }}(\xi )(x_1-x_0) \end{gathered}$$
which is
$\Delta y=f^{{}^{\prime }}(\xi )\Delta x$

suppose an NxM dimensional linear system:
$\vec{y}=[f_1(\vec{x}),f_2(\vec{x}),f_3(\vec{x}),...,f_M(\vec{x})]$

$$\left[\begin{array}{c}\Delta y_0\\ \Delta y_1\\ ...\\ \Delta y_N\end{array}\right]=\left[\begin{array}{c}\frac{\partial f_0(\vec{x})}{\partial x_0},\frac{\partial f_0(\vec{x})}{\partial x_1},\frac{\partial f_0(\vec{x})}{\partial x_2},...,\frac{\partial f_0(\vec{x})}{\partial x_N}\\ \frac{\partial f_1(\vec{x})}{\partial x_0},\frac{\partial f_1(\vec{x})}{\partial x_1},\frac{\partial f_1(\vec{x})}{\partial x_2},...,\frac{\partial f_1(\vec{x})}{\partial x_N}\\ ...\\ \frac{\partial f_M(\vec{x})}{\partial x_0},\frac{\partial f_M(\vec{x})}{\partial x_1},\frac{\partial f_M(\vec{x})}{\partial x_2},...,\frac{\partial f_M(\vec{x})}{\partial x_N}\end{array}\right]\left[\begin{array}{c}\Delta x_0\\ \Delta x_1\\ ...\\ \Delta x_N\end{array}\right]$$
但是这边两个矩阵的形式有点类似，实质上还是有比较大的差别的，一个是状态系数的时间演化，另一个是变量的增量和自变量的增量之间的关系。后者就是一个典型的映射变换，所以需要一个Jacobian。相对应的，海森堡方程的矩阵是一个内涵的。