Assume that a convergent matrix sequence{A<sub>n</sub>}:A<sub>n</sub>→A（n→∞）, A<sub>n</sub>,A∈C<sup>3×3</sup>.We want to form a new matrix sequence {H<sub&...Assume that a convergent matrix sequence{A<sub>n</sub>}:A<sub>n</sub>→A（n→∞）, A<sub>n</sub>,A∈C<sup>3×3</sup>.We want to form a new matrix sequence {H<sub>n</sub>}, derived from {A<sub>n</sub>}, which has also A aslimit and whose convergence is faster than the of {A<sub>n</sub>}. Three rational extrapolation meth-ods for accelerating the convergence of matrix sequences {A<sub>n</sub>} are presented in this paper.The underlying methods are based on the generalized inverse for matrices which is展开更多
This paper proposes a class of generalized mixed least square methods(GMLSM) forthe estimation of weights in the analytic hierarchy process and studies their good properties such asinvariance under transpose, invarian...This paper proposes a class of generalized mixed least square methods(GMLSM) forthe estimation of weights in the analytic hierarchy process and studies their good properties such asinvariance under transpose, invariance under change of scale, and also gives a simple convergent iterativealgorithm and some numerical examples. The well-known eigenvector method(EM) is then compared.Theoretical analysis and the numerical results show that the iterative times of the GMLSM are generallyfewer than that of the MLSM, and the GMLSM are preferable to the EM in several important respects.展开更多
基金The works is supported by the National Natural Science Foundation of China(19871054)
文摘Assume that a convergent matrix sequence{A<sub>n</sub>}:A<sub>n</sub>→A（n→∞）, A<sub>n</sub>,A∈C<sup>3×3</sup>.We want to form a new matrix sequence {H<sub>n</sub>}, derived from {A<sub>n</sub>}, which has also A aslimit and whose convergence is faster than the of {A<sub>n</sub>}. Three rational extrapolation meth-ods for accelerating the convergence of matrix sequences {A<sub>n</sub>} are presented in this paper.The underlying methods are based on the generalized inverse for matrices which is
文摘This paper proposes a class of generalized mixed least square methods(GMLSM) forthe estimation of weights in the analytic hierarchy process and studies their good properties such asinvariance under transpose, invariance under change of scale, and also gives a simple convergent iterativealgorithm and some numerical examples. The well-known eigenvector method(EM) is then compared.Theoretical analysis and the numerical results show that the iterative times of the GMLSM are generallyfewer than that of the MLSM, and the GMLSM are preferable to the EM in several important respects.
