On removal of froissart doublets in Pade - Laplace method

A new algorithm for computing Pade approximants is proposed. It is based on the results concerning the kernel structure of the Toeplitz matrices. Due to the proper choice of a denominator of a Pade approximant, it allows to avoid Froissart doublets appearance in singular blocks of the Pade table. We illustrate the effectiveness of our algorithm through numerical examples.