Speeding Up Network Simulations Using Discrete Time

TL;DR

This paper introduces a discrete time simulation method for network disease spread that accelerates computations by trading some accuracy, supported by theoretical error bounds and cost analysis.

Contribution

It presents a novel discrete time simulation approach for network models, with proven accuracy bounds and an analytical comparison of computational efficiency.

Findings

Discrete time simulation speeds up network disease modeling

Error bounds are established based on numerical methods theory

Computational cost is reduced compared to discrete event simulation

Abstract

We develop a way of simulating disease spread in networks faster at the cost of some accuracy. Instead of a discrete event simulation (DES) we use a discrete time simulation. This aggregates events into time periods. We prove a bound on the accuracy attained. We also discuss the choice of step size and do an analytical comparison of the computational costs. Our error bound concept comes from the theory of numerical methods for SDEs and the basic proof structure comes from the theory of numerical methods for ODEs.