Modifying the BFGS update by a new column scaling technique (Q1338135)

The author proposes a modification of the BFGS method for solving the unconstrained optimization problem `\(\min_{x\in \mathbb{R}^n} f(x)\)', where \(f\in {\mathcal C}^2(\mathbb{R}^n\to \mathbb{R})\). After a brief investigation of the theoretical properties of simple scaling techniques an algorithm that preserves the convergence rate is presented. The results of numerical experiments demonstrate that for ill-conditioned objective functions the advantage of a new algorithm over the unmodified BFGS update can be substantial. An efficient implementation of the new algorithm, which requires \(3n^2+ O(n)\) multiplications per iteration, is proposed.