A Bayesian generative neural network framework for epidemic inference problems.

Area of Science:

• Epidemiology
• Computational Biology
• Network Science

Background:

• Reconstructing epidemic spread on contact networks is crucial for public health interventions.
• Challenges include identifying asymptomatic individuals, the 'patient-zero' problem, and inferring infectivity.
• Exponential growth in possible infection cascades creates significant computational hurdles for accurate inference.

Purpose of the Study:

• To develop a novel framework for reconstructing missing epidemic information on contact networks.
• To address the computational challenges in inferring epidemic cascades compatible with observational data.
• To provide a precise tool for analyzing infectious disease spread in structured populations.

Main Methods:

• A new generative neural network framework was developed.
• The framework learns to generate probable infection cascades that align with observed data.
• The approach is grounded in Bayesian and variational principles.

Main Results:

• The proposed method demonstrated superior or comparable performance against existing techniques.
• Effectiveness was validated across various epidemic inference problems.
• Successful application was shown on both synthetic and real-world contact network data.

Conclusions:

• The generative neural network framework offers a powerful solution for epidemic inference problems.
• Its Bayesian and variational nature allows for high-precision analysis in real-world scenarios.
• The method is particularly promising for understanding infections in workplaces and hospitals.