Beyond Paterson-Stockmeyer: Advancing Matrix Polynomial Computation

[EN] Since 1973, the Paterson¿Stockmeyer method has been considered the most efficient approach for evaluating general matrix polynomials. In this paper, we challenge this long-standing belief by demonstrating that newly developed methods surpass its efficiency. We summarize the state of the art and...

Descripción completa

Detalles Bibliográficos
Autores: Sastre, Jorge|||0000-0002-8612-6717, Ibáñez González, Jacinto Javier|||0000-0002-6912-4453, Alonso Abalos, José Miguel|||0000-0001-6812-7364, Defez Candel, Emilio|||0000-0002-3303-6371
Tipo de recurso: artículo
Fecha de publicación:2025
País:España
Institución:Universitat Politècnica de València (UPV)
Repositorio:RiuNet. Repositorio Institucional de la Universitat Politécnica de Valéncia
Idioma:inglés
OAI Identifier:oai:dnet:riunet______::a288a3b1caaec4634a185da2f4bd68a5
Acceso en línea:https://riunet.upv.es/handle/10251/235361
Access Level:acceso abierto
Palabra clave:Matrix polynomial
Evaluation
Efficient
Stability
Rational
Mixed rational and polynomial
Approximation
Matrix function
13.- Tomar medidas urgentes para combatir el cambio climático y sus efectos
Descripción
Sumario:[EN] Since 1973, the Paterson¿Stockmeyer method has been considered the most efficient approach for evaluating general matrix polynomials. In this paper, we challenge this long-standing belief by demonstrating that newly developed methods surpass its efficiency. We summarize the state of the art and present new results. Additionally, for decades, rational approximations have been deemed superior to polynomial approximations in terms of computational efficiency. However, we reveal that polynomial approximations can achieve a higher order of accuracy than state-of-the-art rational methods at the same computational cost. Through theoretical insights and practical examples, we illustrate the implications of these findings for advanced matrix computations, with potential applications in scientific computing, numerical analysis, and artificial intelligence.