A Divergent Random Walk on Stairs

Abstract: We consider a state-dependent, time-dependent, discrete random walks

X_{t}^{a_{n}}

defined on natural numbers

m a t h b b N

(bent to a "stair" in

m a t h b b N^{2}

) where the random walk depends on input of a positive deterministic sequence

a_{n}

. This walk has the peculiar property that if we set

a_{n}

to be

+ i n f t y

for all

n

, it converges to a stationary distribution

p i (c d o t)

; but if

a_{n}

is uniformly bounded (over all

n

) by any upper bound

a i n (0, i n f t y)

, this walk diverges to infinity with probability 1. It is thus interesting to consider the intermediate case where

a_{n} < i n f t y

for all

n

but

a_{n}

eventually tends to

+ i n f t y

. (Latuszynski et al., 2013) first defined this walk and conjectured that a particular choice of sequence

a_{n}

exists such that (i)

a_{n} o i n f t y

and, (ii)

P (X_{t}^{a_{n}} o i n f t y) = 1

. They managed to construct a sequence

a_{n}

that satisfies (i) and

P (X_{t}^{a_{n}} o i n f t y) > 0

, which is weaker than (ii). In this paper, we obtain a stronger result: for any

s i g m a < 1

, there exists a choice of

a_{n}

so that

P (X_{t} o i n f t y) g e s i g m a

. Our result does not apply when

s i g m a = 1

, the original conjecture remains open. We record our method here for technical interests.

This page was built for publication: A Divergent Random Walk on Stairs