Scalable Calibration of Individual-Based Epidemic Models with Categorical Approximations

A methodological paper by Lorenzo Rimella, published on figshare in 2026, presents a deterministic approach for calibrating individual-based epidemic models. The work includes a real-world application to the 2001 UK Foot-and-Mouth outbreak, incorporating data from 162,775 farms. The method uses categorical distributions to approximate likelihoods, enabling parameter estimation with automatic differentiation libraries like TensorFlow.

Use Cases

• Calibrating individual-based epidemic models based on the described categorical approximation method.
• Estimating parameters for models with individual-specific transition rates based on the automatic differentiation approach.
• Analyzing disease outbreaks with spatial interactions and under-reporting using the scalable likelihood framework.
• Comparing the performance of approximate inference methods against sampling-based competitors as demonstrated in the paper.

Strengths

• Includes a real-world application to a major outbreak, modeling 162,775 farms.
• Method is designed for scalability and reduced computational cost compared to competitor methods.
• Licensed under CC-BY-4.0 for open sharing and reuse.

Limitations

• The dataset is a 5.6 MB PDF document; the underlying data files are not directly provided.
• Column-level documentation and sample data are unavailable, limiting immediate analysis.
• Row count is unknown, which may limit suitability assessment for direct ML application.

Provenance

Source

Lorenzo Rimella via figshare.

Collection Method

Likely contains results from applying a novel calibration method to synthetic and real-world epidemic data.

Time Range

Includes analysis of the 2001 UK Foot-and-Mouth outbreak.

Freshness

Last updated 2026-05-11 13:56:22; freshness should be verified.

Geography

United Kingdom (for the Foot-and-Mouth disease application).