背景
对一个多元函数 f ( x ) f(x) f(x) 求最小值,当无法准确求出其准确结果时,需要用到其导数。
根据泰勒公式, f ( x ) f(x) f(x) 在 x k x_k xk 处展开二阶导:
f ( x ) ≈ f ( x k ) + ∇ x f ′ ( x k ) ( x − x k ) T + 1 2 ( x − x k ) T ∇ x 2 f ′ ′ ( x k ) ( x − x k ) f(x) \approx f(x_k) + \nabla_x f'(x_k)(x - x_k)^T + \frac{1}{2} (x - x_k)^T \nabla_x^2f''(x_k) (x - x_k) f(x)≈f(xk)+∇xf′(xk)(x−xk)T+21(x−xk)T∇x2f′′(xk)(x−xk)
其中,一阶导梯度和二阶导 H e s s i a n Hessian Hessian 矩阵如下:
g k = f ′ ( x k ) = ( ∂ f ( x k ) ∂ x 1 , ∂ f ( x k ) ∂ x 2 , … , ∂ f ( x k ) ∂ x n ) H k − 1 = f ′ ′ ( x k ) − 1 = ( ∂ 2 f ( x k ) ∂ 2 x 1 2 ⋯ ∂ 2 f ( x k ) ∂ x 1 ∂ x n ⋮ ∂ 2 f ( x k ) ∂ x i ∂ x j ⋮ ∂ 2 f ( x k ) ∂ x n ∂ x 1 ⋯ ∂