An Extension of Two Conjugate Direction Methods to Markov Chain Problems

keywords: Krylov subspace methods, conjugate direction methods, Markov break chains, stationary probability distribution
Motivated by the recent applications of the conjugate residual method to nonsymmetric linear systems by Sogabe, Sugihara and Zhang [An extension of the conjugate residual method to nonsymmetric linear systems. J. Comput. Appl. Math., Vol. 266, 2009, pp. 103--113], this paper describes two conjugate direction methods, BiCR and BiCG, and attempts to extend their applications to compute the stationary probability distribution for an irreducible Markov chain with the aim of finding an alternative basic solver. Numerical experiments show the feasibility of the BiCR and BiCG to some extent, with applications to several practical Markov chain problems.
mathematics subject classification 2000: 60J22, 65C40
reference: Vol. 34, 2015, No. 2, pp. 495–516