Multitask learning and nonlinear optimal control of the COVID-19 outbreak: A geometric programming approach

Area of Science:

• Epidemiology
• Computational modeling
• Public health policy

Background:

• Compartmental epidemic models are crucial for understanding disease spread.
• Integrating human mobility data enhances epidemic modeling accuracy.
• Optimizing non-pharmaceutical interventions requires balancing health and economic outcomes.

Purpose of the Study:

• To develop a multitask learning framework for parameter estimation in a discrete-time epidemic model.
• To design optimal human mobility restriction strategies that minimize epidemic spread and economic costs.
• To provide a computationally efficient method for solving the proposed optimal control problem.

Main Methods:

• An extended SEIR (Susceptible-Exposed-Infectious-Recovered) model incorporating human mobility dynamics.
• Multitask learning utilizing regional death data and cellphone-based mobility patterns.
• Nonlinear optimal control formulation solved via geometric programming.
• A practical approach for setting economic cost budgets based on hospital capacity.

Main Results:

• The multitask learning approach effectively learns epidemic model parameters from diverse data sources.
• Optimal control strategies were derived to curb disease spread while adhering to economic constraints.
• Geometric programming provides an efficient, polynomial-time solution for the control problem.
• The method successfully simulated COVID-19 pandemic data from the Philadelphia area.

Conclusions:

• The proposed framework offers a robust method for data-driven epidemic modeling and control.
• Optimizing mobility restrictions can effectively mitigate epidemics with minimized economic disruption.
• The approach provides a valuable tool for policymakers in managing public health crises.