Optimize Data-Driven Multi-Agent Simulation for COVID-19 Transmission

Abstract

Background: Multi-Agent Simulation is an essential technique for exploring complex systems. In researches of contagious diseases, it is widely exploited to analyze their spread mechanisms, especially for preventing COVID-19. Nowadays, transmission dynamics and interventions of COVID-19 have been elaborately established by this method, but its computation performance is seldomly concerned. As it usually suffers from inadequate CPU utilization and pour data locality, optimizing the performance is challenging. Results: This paper explores approaches to optimize multi-agent simulation for COVID-19 disease. The focus of this work is on the algorithm and data structure designs for improving performance, as well as its parallelisation strategies. We propose two successive methods to optimize the computation. We construct a case-focused iteration algorithm to improve data locality, and create a thread-safe data-mapping paradigm called hierachical hash table to accelerate hash operations. Conclusions: Our performance results demonstrate capabilities of these methods exhibiting significant improvements of system performance. The case-focused method degrades $\sim 90 \%$ cache references and achieves $\times 4.3$ speedup. Hierachical hash table can further boost computation speed by 47\%. And parallel implementation with 20 threads on CPU achieves $\times 81$ speedup consequently.

COVID-19 Modeling: A Review

“…It is also important to infer their epidemiological attributes and to understand how the coronavirus spreads spatially, temporally, and socially [35, 34]. In this regard, intensive COVID-19 modeling research focuses on exploring the source and spectrum of the COVID-19 infection, identifying clinical and epidemiological characteristics, tracking transmission routes, and forecasting case development trends and the peak values of infected cases and disease transmission [35, 34, 22, 50, 293, 294, 95, 125, 184]. The related work aim to understand the characteristics and dynamics of the coronavirus and COVID-19 disease to inform disease precaution, characterize the epidemiological attributes of coronavirus and COVID-19 and their resulting infections, mortality and patient statistics, quantifying the influence of virus containment and mitigation campaigns on epidemic dynamics and virus transmission, and measuring the influence of epidemic and infections on medical resource planning, etc.…”

“…The related work aim to understand the characteristics and dynamics of the coronavirus and COVID-19 disease to inform disease precaution, characterize the epidemiological attributes of coronavirus and COVID-19 and their resulting infections, mortality and patient statistics, quantifying the influence of virus containment and mitigation campaigns on epidemic dynamics and virus transmission, and measuring the influence of epidemic and infections on medical resource planning, etc. Typical modeling methods include mathematical and statistical models such as linear and nonlinear regression models [22, 321, 184], compartmental models including those incorporated with statistical settings and customization such as time and age dependence, SIR with Markov chain Monte Carlo [321], simulation methods [52, 125], and network modeling [95, 167].…”