CSonNet: An Agent-Based Modeling Software System for Discrete Time Simulation
Contagion dynamics on networks are used to study many problems, including disease and virus epidemics, incarceration, obesity, protests and rebellions, needle sharing in drug use, and hurricane and other natural disaster events. Simulators to study these problems range from smaller-scale serial codes to large-scale distributed systems. In recent years, Python-based simulation systems have been built. In this work, we describe a new Python-based agent-based simulator called CSonNet. It differs from codes such as Epidemics on Networks in that it performs discrete time simulations based on the graph dynamical systems formalism. CSonNet is a parallel code; it implements concurrency through an embarrassingly parallel approach of running multiple simulation instances on a user-specified number of forked processes. It has a modeling framework whereby agent models are composed using a set of pre-defined state transition rules. We provide strong-scaling performance results and case studies to illustrate its features.
Authors:
; ; ; ; ;
Award ID(s):
Publication Date:
NSF-PAR ID:
10376931
Journal Name:
2021 Winter Simulation Conference (WSC)
Page Range or eLocation-ID:
1 to 12
4. We study a distributed policy evaluation problem in which a group of agents with jointly observed states and private local actions and rewards collaborate to learn the value function of a given policy via local computation and communication. This problem arises in various large-scale multi-agent systems, including power grids, intelligent transportation systems, wireless sensor networks, and multi-agent robotics. We develop and analyze a new distributed temporal-difference learning algorithm that minimizes the mean-square projected Bellman error. Our approach is based on a stochastic primal-dual method and we improve the best-known convergence rate from $O(1/\sqrt{T})$ to $O(1/T)$, where $T$ is the total number of iterations. Our analysis explicitly takes into account the Markovian nature of the sampling and addresses a broader class of problems than the commonly-used i.i.d. sampling scenario.