Abstract
A great potential exists for mitigating the computational costs of spatially
explicit agent-based models (SE-ABMs) by taking advantage of parallel and
high-performance computing. However, spatial dependency and heterogeneity of
interactions between agents pose challenges for parallel SE-ABMs to achieve good scalability. This chapter summarizes an application of the principle of data locality to tackle these challenges by extending a theoretical approach to the representation of the spatial computational domain. We propose and formalize a graph-based locality-aware approach to scalable parallelization of SE-ABMs. To demonstrate the applicability of this approach, two sets of experimentation are laid out and a locality-aware algorithm is designed to facilitate the study of model scalability. The results of simulation experiments illustrate the advantage of our approach to scalable parallel agent-based models of spatial interaction.
explicit agent-based models (SE-ABMs) by taking advantage of parallel and
high-performance computing. However, spatial dependency and heterogeneity of
interactions between agents pose challenges for parallel SE-ABMs to achieve good scalability. This chapter summarizes an application of the principle of data locality to tackle these challenges by extending a theoretical approach to the representation of the spatial computational domain. We propose and formalize a graph-based locality-aware approach to scalable parallelization of SE-ABMs. To demonstrate the applicability of this approach, two sets of experimentation are laid out and a locality-aware algorithm is designed to facilitate the study of model scalability. The results of simulation experiments illustrate the advantage of our approach to scalable parallel agent-based models of spatial interaction.
Original language | English |
---|---|
Title of host publication | Advances in Geocomputation: |
Subtitle of host publication | Geocomputation 2015--The 13th International Conference |
Editors | Daniel A. Griffith, Yongwan Chun, Denis J. Dean |
Publisher | Springer |
Pages | 405-423 |
ISBN (Electronic) | 9783319227863 |
ISBN (Print) | 9783319227856 |
DOIs | |
Publication status | Published - 5 Jan 2017 |
Event | GeoComputation 2015: The 13th International Conference - University of Texas at Dallas, Dallas, Texas, United States Duration: 20 May 2015 → 23 May 2015 http://www.utdallas.edu/geocomputation/ |
Publication series
Name | Advances in Geographic Information Science |
---|---|
Publisher | Springer |
ISSN (Print) | 1867-2434 |
ISSN (Electronic) | 1867-2442 |
Conference
Conference | GeoComputation 2015 |
---|---|
Country/Territory | United States |
City | Dallas, Texas |
Period | 20/05/15 → 23/05/15 |
Internet address |
Keywords
- Locality awareness
- Parallel agent-based models
- Spatial interaction