Cliff Kerr - Starsim: A flexible framework for agent-based modeling of health and disease

StarSim is a flexible agent-based modeling framework for health and disease simulation, designed to be both fast and user-friendly

The framework uses array-based implementation instead of traditional object-oriented approach, making it up to 100x faster than conventional agent-based models

Can simulate multiple diseases simultaneously and model disease interactions (like HIV-syphilis co-infection), which most other models cannot do

Supports networks (household, school, workplace, community) and allows agents to interact within these networks

Built with Python using NumPy arrays for performance, with each agent being an entry in the array rather than a separate object

Can handle large-scale simulations of up to 1 billion agents, with each person taking about 10KB of memory

Includes built-in calibration tools based on the Optuno library and supports parallel processing

Designed to be accessible to users in developing countries where disease burden is highest, not just advanced users in global north

Uses Cyrus library to simplify common scientific computing tasks and provide sensible defaults

Originally developed for COVID-19 modeling but expanded to include various diseases with different timescales and transmission patterns

Open-source and modular design allows users to customize and extend the framework for different diseases or modeling needs