DAPT: A package enabling distributed automated parameter testing.

Area of Science:

• Computational Science
• Simulation Modeling
• Scientific Computing

Background:

• Agent-based models (ABM) and simulations necessitate extensive parameter testing for robust evaluation.
• Managing large-scale parameter sweeps and simulation data presents significant computational and logistical challenges.
• Current methods for parameter testing and analysis can be inefficient, especially when distributed across teams or high-performance computing (HPC) resources.

Purpose of the Study:

• To introduce the Distributed Automated Parameter Testing (DAPT) Python package.
• To provide a flexible and scriptable toolset for managing and executing parameter testing for simulation models.
• To enable efficient utilization of distributed computational resources, including crowdsourcing, for model evaluation.

Main Methods:

• Development of the DAPT Python package for automated parameter testing.
• Implementation of a distributed approach using an online parameter database.
• Facilitation of simultaneous parameter set execution across multiple users and computational resources.
• Integration of flexible, scriptable tools for job management and data processing.

Main Results:

• DAPT enables simultaneous execution of parameter sets from multiple distributed sources.
• The package facilitates ad hoc crowdsourcing of computational power for parameter sweeps.
• DAPT streamlines the management and storage of simulation data.
• The toolset allows for rapid evaluation of models and hypotheses.

Conclusions:

• The DAPT Python package offers an effective solution for accelerating parameter testing in agent-based models and simulations.
• Distributed and crowdsourced computation via DAPT enhances the efficiency of scientific model evaluation.
• DAPT provides a flexible framework for teams to manage complex simulation testing and analysis workflows.