Efficient computational method for matrix function in dynamic problems.

An algorithm based on the Paterson-Stockmeyer (PS) scheme and filtering technology is developed to compute the large matrix functions in dynamic problems accurately and efficiently. With the assistance of analysis on truncation error and error caused by filtering, the error growth law during the computation is studied, based on which an adaptive filtering threshold is proposed to help the proposed algorithm more efficiently achieve similar accuracy as the original PS scheme. Numerical examples, including 30 random matrices with different bandwidths, 10 adjacency matrices in the complex network dynamic problems, and a trampoline vibration problem, are given to verify the efficiency and accuracy of the proposed algorithm. Numerical results suggest that the proposed algorithm can achieve good accuracy and efficiency in computing the matrix function in the considered dynamic problems. [ABSTRACT FROM AUTHOR]