Skip to content

Latest commit

 

History

History
25 lines (13 loc) · 3.73 KB

README.md

File metadata and controls

25 lines (13 loc) · 3.73 KB

PEP8 Pytest GitHub

Particle Swarm Optimization for Calibration in Spatially Explicit Agent-Based Modeling

Authors: Alexander Michels, Jeon-Young Kang, Shaowen Wang

Special thanks to Zhiyu Li and Rebecca Vandewalle for suggestions and feedback on this notebook!

Open with CyberGISX

Gif of Particle Swarm Optimization

This notebook is related to an upcoming publication entitled "Particle Swarm Optimization for Calibration in Spatially Explicit Agent-Based Modeling." The abstract for the paper is:

A challenge in computational modeling of complex geospatial systems is the amount of time and resources required to tune a set of parameters that reproduces the observed patterns of phenomena of being modeled. Well-tuned parameters are necessary for models to reproduce real-world multi-scale space-time patterns, but calibration is often computationally-intensive and time-consuming. Particle Swarm Optimization (PSO) is a swarm intelligence optimization algorithm that has found wide use for complex optimization including non-convex and noisy problems. In this study, we propose to use PSO for calibrating parameters in spatially explicit agent-based models (ABMs). We use a spatially explicit ABM of influenza transmission based in Miami, Florida, USA as a case study. Further, we demonstrate that a standard implementation of PSO can be used out-of-the-box to successfully calibrate models and out-performs Monte Carlo in terms of optimization and efficiency.

The notebook is designed to teach you about Particle Swarm Optimization (PSO) and how you can use it for parameter optimization. Particle Swarm Optimization (PSO) was first introduced in 1995 by James Kennedy and Russell Eberhart. The algorithm began as a simulation of social flocking behaviors like those exhibited by flocks of birds and schools of fish, specifically of birds searching a cornfield, but was found to be useful for training feedforward multilayer pernceptron neural networks. Since then, PSO has been adapted in a variety of ways and applied to problems including wireless-sensor networks, classifying biological data, scheduling workflow applications in cloud computing environments, Image classification and power systems. In this notebook we explore PSO's usefulness for calibration, with a focus on spatially-explicit agent-based models (ABMs).

The model is also available on CoMSES, you can find it by clicking on the badge below:

Open Code Badge