A new self-scaling memoryless quasi-Newton update for unconstrained optimization

Based on the augmented version of the quasi-Newton method proposed by Aminifard et al. (App. Num. Math. 167:187–201, 2021), a new scaled parameter of the self-scaling memoryless BFGS update formula is proposed. The idea is to cluster the eigenvalues of the search direction matrix, obtained by minimizing the difference between the largest and the smallest eigenvalues of the matrix. The sufficient descent property is proved for uniformly convex functions, and the global convergence of the proposed algorithm is proved both for the uniformly convex and general nonlinear objective functions. Numerical experiments on a set of test functions of the CUTEr collection show that the proposed method is efficient. In addition, the proposed algorithm is effectively applied to salt and pepper noise elimination problem.