Exhibits, Demos & Posters
Parameter Estimation using Scientific Workflows
- David Abramson, Faculty of Information Technology, Monash University, Victoria, Australia
- Colin Enticott, Faculty of Information Technology, Monash University, Victoria, Australia
- Tom Peachey, Faculty of Information Technology, Monash University, Victoria, Australia
Parameter estimation techniques allow the exploration of the design space of a computational model. In recent years there has been interest in performing parameter sweeps and searches across "scientific workflows," however, many existing workflow tools are not well suited to this. In this poster we show that is possible to augment existing systems with a small set of special "actors" that implement the parameter estimation logic. Specifically, we discuss a set of new Kepler actors that support both complete and partial sweeps based on experimental design techniques. When combined with a novel parallel execution mechanism, we are able to execute parallel sweeps and searches across workflows that run on distributed "grid" infrastructure. We illustrate our new system with a case study in cardiac cell modeling.