## On the numerical analysis of infinite multi-dimensional Markov chains

##### Author

Orhan, Muhsin Can

##### Advisor

Dayar, Tuğrul

##### Date

2017-08##### Publisher

Bilkent University

##### Language

English

##### Type

Thesis##### Item Usage Stats

41

views

views

0

downloads

downloads

##### Metadata

Show full item record##### Abstract

A system with multiple interacting subsystems that exhibits the Markov property
can be represented as a multi-dimensional Markov chain (MC). Usually the
reachable state space of this MC is a proper subset of its product state space,
that is, Cartesian product of its subsystem state spaces. Compact storage of the
infinitesimal generator matrix underlying such a MC and efficient implementation
of analysis methods using Kronecker operations require the set of reachable
states to be represented as a union of Cartesian products of subsets of subsystem
state spaces.
We first show that the problem of partitioning the reachable state space of a
three or higher dimensional system with a minimum number of partitions into
Cartesian products of subsets of subsystem state spaces is NP-complete. Two
algorithms, one merge based the other refinement based, that yield possibly nonoptimal
partitionings are presented. Results of experiments on a set of problems
from the literature and those that are randomly generated indicate that, although
it may be more time and memory consuming, the refinement based algorithm
almost always computes partitionings with a smaller number of partitions than
the merge based algorithm. When the infinitesimal generator matrix underlying the MC is represented
compactly using Kronecker products, analysis methods based on vector–
Kronecker product multiplication need to be employed. When the factors in the
Kronecker product terms are relatively dense, vector–Kronecker product multiplication
can be performed efficiently by the shuffle algorithm. When the factors are
relatively sparse, it may be more efficient to obtain nonzero elements of the generator
matrix in Kronecker form on-the-fly and multiply them with corresponding
elements of the vector. We propose a modification to the shuffle algorithm that
multiplies relevant elements of the vector with submatrices of factors in which zero
rows and columns are omitted. This approach avoids unnecessary floating-point operations that evaluate to zero during the course of the multiplication. Numerical
experiments on a large number of models indicate that, in many cases the
modified shuffle algorithm performs a smaller number of floating-point operations
than the shuffle algorithm and the algorithm that generates nonzeros on-the-fly,
sometimes with minimum number of floating-point operations and amount of
memory possible.
Although the generator matrix is stored compactly using Kronecker products,
solution vectors used in the analysis still require memory proportional to the
size of the reachable state space. This becomes a bigger problem as the number
of dimensions increases. We show that it is possible to use the hierarchical
Tucker decomposition (HTD) to store the solution vectors during Kroneckerbased
Markovian analysis relatively compactly and still carry out the basic operation
of vector–matrix multiplication in Kronecker form relatively efficiently. The time evolution of a stochastic chemical system modelled as a continuoustime
MC (CTMC) can be described as a system of ordinary differential equations
(ODEs) known as the chemical master equation (CME). The CME can be analyzed
by discretizing time and solving a linear system obtained by truncating
the countably infinite state space at each time step. However, it is not trivial to
choose a truncated state space that includes few states with negligible probabilities
and leaves out only a small probability mass. We show that it is possible to
decrease the memory requirement of the ODE solver using HTD with adaptive
truncation strategies and we propose a novel approach to truncate the countably
infinite state space using prediction vectors. Numerical experiments indicate that
adaptive truncation strategies improve time and memory efficiency significantly
when fixed truncation strategies are inefficient. Finally, we consider a multi-class multi-server retrial queueing system in
which customers arrive according to a class-dependent Markovian arrival process
(MAP). Service and retrial times follow class-dependent phase-type (PH)
distributions with the further assumption that PH distributions of retrial times
are acyclic. Here, we obtain a necessary and sufficient condition for ergodicity
from criteria based on drifts. The countably infinite state space of the model
is truncated with an appropriately chosen Lyapunov function. The truncated
model is described as a multi-dimensional MC and a Kronecker representation of
its infinitesimal generator matrix is numerically analyzed.

##### Keywords

Markov chainKronecker representation

Cartesian product partitioning

Vector–Kronecker product multiplication

Hierarchical Tucker decomposition

Retrial queues.

Chemical master equation