Leveraging deep generative modelling at the agent level in discrete-event simulations
Abstract
Emulators are often invoked to facilitate parameter inference in computationally intensive agent-based simulations (ABSs). Most approaches, however, require a large sample of pre-existing outputs on which such emulators can be trained. For computationally intensive simulations, this sample can be prohibitively expensive to generate. We propose a new approach to leverage emulators at the agent level in event-based ABSs to reduce the computational requirements of the simulation itself. We demonstrate this in a computationally intensive ``multi-disease'' ABS, Thanzi La Onse, which self-consistently simulates a wide range of medical conditions in a population. By applying conditional tabular generative adversarial networks to replace the modelling of one of the simulated causes of ill health, road traffic injuries (RTIs), we are able to improve its simulation run-time by 86.6% (95% bootstrap confidence interval [85.2, 88.1]), while recovering the evolution of cause-specific mean disability adjusted life years with a 13% error, despite the low assumed incidence of fatal RTIs (~2%). Crucially, we do so while retaining the full ABS framework, and correctly recovering multi-disease dynamics. This represents a crucial step towards enabling full-model emulators for very computationally intensive simulations, and making large ABS amenable to rigorous parameter inference and a full understanding of parameter sensitivity.
Related
- Score-Based Generative Data Assimilation for Integrating Aggregated Surveillance Data into Agent-Based Models in Epidemic Tracking
- Limited knowledge exchange across biological subfields in agent-based modelling: an LLM-assisted bibliometric analysis
- Agent‐Based Modelling of Microbialite Formation Through Sedimentation and Precipitation Dynamics
- Leveraging prompt-driven generative AI for systematic reviews in digital psychiatry: A stage-matched comparative proof-of-concept for healthcare researchers and clinicians
- Modelling sexual partnership dynamics and population heterogeneities in agent-based dynamic network models
BibTeX
@misc{molaro2026leveraging,
title = {{Leveraging deep generative modelling at the agent level in discrete-event simulations}},
author = {Margherita Molaro and Matthew M. Graham and Oliver Watson and Rachel E. Murray-Watson and Martin Chalkley and Joseph H. Collins and Bingling She and Asif U. Tamuri and Timothy B. Hallett},
howpublished = {arXiv},
year = {2026},
month = nov,
doi = {10.48550/arxiv.2611.07995},
eprint = {2611.07995},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2611.07995},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Leveraging deep generative modelling at the agent level in discrete-event simulations
AU - Margherita Molaro
AU - Matthew M. Graham
AU - Oliver Watson
AU - Rachel E. Murray-Watson
AU - Martin Chalkley
AU - Joseph H. Collins
AU - Bingling She
AU - Asif U. Tamuri
AU - Timothy B. Hallett
PY - 2026
DA - 2026/11/05
PB - arXiv
DO - 10.48550/arxiv.2611.07995
UR - https://arxiv.org/abs/2611.07995
AB - Emulators are often invoked to facilitate parameter inference in computationally intensive agent-based simulations (ABSs). Most approaches, however, require a large sample of pre-existing outputs on which such emulators can be trained. For computationally intensive simulations, this sample can be prohibitively expensive to generate. We propose a new approach to leverage emulators at the agent level in event-based ABSs to reduce the computational requirements of the simulation itself. We demonstrate this in a computationally intensive ``multi-disease'' ABS, Thanzi La Onse, which self-consistently simulates a wide range of medical conditions in a population. By applying conditional tabular generative adversarial networks to replace the modelling of one of the simulated causes of ill health, road traffic injuries (RTIs), we are able to improve its simulation run-time by 86.6% (95% bootstrap confidence interval [85.2, 88.1]), while recovering the evolution of cause-specific mean disability adjusted life years with a 13% error, despite the low assumed incidence of fatal RTIs (~2%). Crucially, we do so while retaining the full ABS framework, and correctly recovering multi-disease dynamics. This represents a crucial step towards enabling full-model emulators for very computationally intensive simulations, and making large ABS amenable to rigorous parameter inference and a full understanding of parameter sensitivity.
ER -