MarketDavidon–Fletcher–Powell formula
Company Profile

Davidon–Fletcher–Powell formula

The Davidon–Fletcher–Powell formula finds the solution to the secant equation that is closest to the current estimate and satisfies the curvature condition. It was the first quasi-Newton method to generalize the secant method to a multidimensional problem. This update maintains the symmetry and positive definiteness of the Hessian matrix.

Compact representation
By unwinding the matrix recurrence for B_k , the DFP formula can be expressed as a compact matrix representation. Specifically, defining S_k = \begin{bmatrix} s_0 & s_1 & \ldots & s_{k-1} \end{bmatrix}, Y_k = \begin{bmatrix} y_0 & y_1 & \ldots & y_{k-1} \end{bmatrix}, and upper triangular and diagonal matrices \big(R_k\big)_{ij} := \big(R^{\text{SY}}_k\big)_{ij} = s^T_{i-1}y_{j-1}, \quad \big(R^{\text{YS}}_k\big)_{ij} = y^T_{i-1}s_{j-1}, \quad (D_k)_{ii} := \big(D^{\text{SY}}_k\big)_{ii} = s^T_{i-1}y_{i-1} \quad \quad \text{ for } 1 \le i \le j \le k the DFP matrix has the equivalent formula B_k = B_0 + J_k N^{-1}_k J^T_k, J_k = \begin{bmatrix} Y_k & Y_k - B_0 S_k \end{bmatrix} N_k = \begin{bmatrix} 0_{k \times k} & R^{\text{YS}}_k \\ \big( R^{\text{YS}}_k \big)^T & R_k+R^T_k-(D_k+S^T_k B_0 S_k) \end{bmatrix} The inverse compact representation can be found by applying the Sherman-Morrison-Woodbury inverse to B_k. The compact representation is particularly useful for limited-memory and constrained problems. ==See also==
tickerdossier.comtickerdossier.substack.com