Computable bounds of an ^2-spectral gap for discrete Markov chains with band transition matrices

Abstract: We analyse the

e l l^{2} (p i)

-convergence rate of irreducible and aperiodic Markov chains with

N

-band transition probability matrix

P

and with invariant distribution

p i

. This analysis is heavily based on: first the study of the essential spectral radius

r_{e s s} (P_{| e l l^{2} (p i)})

of

P_{| e l l^{2} (p i)}

derived from Hennion's quasi-compactness criteria; second the connection between the Spectral Gap property (SG

_{2}

) of

P

on

e l l^{2} (p i)

and the

V

-geometric ergodicity of

P

. Specifically, (SG

_{2}

) is shown to hold under the condition

a l p h a_{0} := s u m_{m = - N}^{N} l i m s u p_{i i g h t a r r o w + i n f t y} s q r t P (i, i + m), P^{*} (i + m, i) e x t l e s s, 1

Moreover

r_{e s s} (P_{| e l l^{2} (p i)}) l e q a l p h a_{0}

. Effective bounds on the convergence rate can be provided from a truncation procedure.

Recommendations

Cited in

(3)

This page was built for publication: Computable bounds of an \(\ell^2\)-spectral gap for discrete Markov chains with band transition matrices