RLS算法-公式推导
不带遗忘因子的推导:递推最小二乘法推导(RLS)——全网最简单易懂的推导过程 - 阿Q在江湖的文章 - 知乎
https://zhuanlan.zhihu.com/p/111758532
对于一组观测点 ( x 1 , y 1 ) (x_1, y_1) (x1,y1), ( x 2 , y 2 ) (x_2, y_2) (x2,y2), ⋯ \cdots ⋯, ( x n , y n ) (x_n, y_n) (xn,yn),用 y ^ = k x + b \hat{y} = kx + b y^=kx+b进行线性拟合,有如下优化问题:
e r r m i n = m i n ∑ i = 1 n ζ n − i ( y ^ − y i ) 2 = m i n ∑ i = 1 n ζ n − i ( k x i + b − y i ) 2 err_{min} = min \sum_{i=1}^n \zeta^{n - i} (\hat{y} - y_i)^2 =min \sum_{i=1}^n \zeta^{n - i} (kx_i + b - y_i)^2 errmin=mini=1∑nζn−i(y^−yi)2=mini=1∑nζn−i(kxi+b−yi)2
记 f ( k , b ) = ∑ i = 1 n ζ n − i ( k x i + b − y i ) 2 f(k,b) = \sum_{i=1}^n \zeta^{n - i} (kx_i + b - y_i)^2 f(k,b)=∑i=1nζn−i(kxi+b−yi)2,分别对 k , b k,b k,b求偏导,令其等于 0 0 0,有
{ ∂ f ∂ k = ∑ i = 1 n ζ n − i ( k x i + b − y i ) x i = 0 ∂ f ∂ b = ∑ i = 1 n ζ n − i ( k x i + b − y i ) = 0 \begin{cases} \frac {\partial f } {\partial k } = \sum_{i=1}^n \zeta^{n - i} (kx_i + b - y_i)x_i = 0 \\ \quad \\ \frac {\partial f } {\partial b } = \sum_{i=1}^n \zeta^{n - i} (kx_i + b - y_i) = 0 \end{cases} ⎩⎪⎨⎪⎧∂k∂f=∑i=1nζn−i(kxi+b−yi)xi=0∂b∂f=∑i=1nζn−i(kxi+b−yi)=0
改写成矩阵形式:
( ∑ i = 1 n ζ n − i x i 2 ∑ i = 1 n ζ n − i x i ∑ i = 1 n ζ n − i x i ∑ i = 1 n ζ n − i ) ∗ ( k b ) = ( ∑ i = 1 n ζ n − i x i y i ∑ i = 1 n ζ n − i y i ) (1) \begin{pmatrix}\sum_{i=1}^n\zeta^{n-i}x_i^2 & \sum_{i=1}^n\zeta^{n-i}x_i \\ \quad \\ \sum_{i=1}^n\zeta^{n-i}x_i & \sum_{i=1}^n\zeta^{n-i} \end{pmatrix} * \begin{pmatrix} k \\ \quad \\ b \end{pmatrix} = \begin{pmatrix} \sum_{i=1}^n\zeta^{n-i}x_iy_i \\ \quad \\ \sum_{i=1}^n\zeta^{n-i}y_i \end{pmatrix} \tag{1} ⎝⎛∑i