Osgood, Nathaniel, "Lightening the Performance Burden of Individual-Based Models through Dimensional Analysis and Scale Modeling", 2007 July 29-2007 August 2
While individual-based models are attractive for addressing certain types of modeling problems, such models impose (frequently dramatically) higher performance costs for larger populations. Lengthy simulation times inhibit interactive learning, and given limited modeler time can impose higher opportunity costs by limiting model comprehension, refinement and user interaction. This paper proposes the novel use of dimensional analysis and scale modeling which have long played an important role in understanding physical systems to lessen the performance barriers associated with simulation of individual-based models. Given a dimensionally homogeneous (full-scale) simulation model with a large population, we propose a precise, rigorous, systematic and general-purpose technique to formulate a reduced-scale individual-based model that simulates a smaller population. Measurements made of particular output parameters of such reduced-scale models can then be precisely transformed (in accordance with model scaling laws) to yield comparable results for a full-scale model without the need to run the full-scale model. While discretization effects limit the degree of scaling that can be achieved, these techniques are notable in relying only upon dimensional homogeneity of the full-scale model, and on not the specifics of model behavior or use of a particular mathematical framework.
This is the whole item.
Date created |
- 2007 July 29-2007 August 2
|
Type |
|
Processing Activity |
|
License |
|