Pasar al contenido principal

Resultados relacionados

Resultados totales (Incluyendo duplicados): 2
Encontrada(s) 1 página(s)

RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

High-level convergence order accelerators of iterative methods for nonlinear problems

  • Cordero Barbero, Alicia|||0000-0002-7462-9173
  • Torregrosa Sánchez, Juan Ramón|||0000-0002-9893-0761
  • Rojas-Hiciano, Renso V.
  • Vassileva, María P.
[EN] We present an efficient strategy to increase, under certain conditions, the order of convergence of iterative methods to solve nonlinear systems of equations. We analytically compare the new accelerator with others and establish the conditions under which this technique is more efficient. We perform an analysis of the efficiency of some one-step accelerators that increase the convergence order by two units. New concepts about efficiency are introduced which allow us to compare different iterative schemes from other points of view. We demonstrate that our proposal is a good alternative to the existing ones. As a consequence, we propose two new maximally efficient, damped Newton-Traub type schemes of order 5 and 6. These are an improvement of two other maximally efficient methods. Their numerical performance is better than that of known methods of the same order, and we find that it is a very economical way to achieve high order. Some numerical examples confirm the theoretical results., This research was partially financed by Fondo Nacional de Innovacion y Desarrollo Cientifico y Tecnologico (FONDOCYT) of Ministerio de Educacion Superior Ciencia y Tecnologia de la Republica Dominicana (MESCYT) grant number FONDOCYT 2023-1-1D2-0537.
RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Publicación Artículo científico (JournalArticle)

Efficiency and Stability of a New Hybrid Unconstrained Optimization Algorithm with Quasi-Newton Updates and Higher-Order Methods

  • Cordero Barbero, Alicia|||0000-0002-7462-9173
  • Torregrosa Sánchez, Juan Ramón|||0000-0002-9893-0761
  • Maimo, Javier G.
  • Ureña Castillo, Natanael
[EN] We propose the higher-order quasi-Newton (HOQN) method, a hybrid algorithm for unconstrained optimization that combines Newtonian predictors with higher-order correctors derived from vector extensions of the Traub, Chun, and Ostrowski methods, along with quasi-Newton updates of the inverse Hessian using Broyden-Fletcher-Goldfarb-Shanno (BFGS) or Davidon-Fletcher-Powell (DFP) formulas. We demonstrate that the resulting scheme achieves cubic local convergence order, representing a substantial improvement over the superlinear convergence typical of classical quasi-Newton methods, while maintaining a cost of On2 per iteration. We also analyze variants that incorporate two successive quasi-Newton updates, and show that they retain the same cubic order. Numerical experiments with the benchmark functions of Himmelblau and Freudenstein-Roth confirm the theoretical convergence order and show that the hybrid variants consistently require fewer iterations than BFGS, DFP, and Symmetric Rank-One (SR1). In the case of the Booth function, given its strictly convex quadratic structure, the proposed hybrid methods reach the global minimum in just two iterations and exhibit numerical accuracy superior to that of classical quasi-Newton methods. In addition, limited-memory variants (L-HOQN) are introduced; these are evaluated during the training of a convolutional neural network on the MNIST dataset, where they achieve test accuracies exceeding 99% and outperform L-BFGS and standard stochastic gradient descent (SGD) at all tested learning rates., This research was funded by the Fondo Nacional de Innovacion y Desarrollo Cientifico y Tecnologico (FONDOCYT) of the Ministerio de Educacion Superior, Ciencia y Tecnologia de la Republica Dominicana (MESCyT), under grant number FONDOCYT 2023-1-1D2-0537.