Policy-Embedded Graph Expansion: Networked HIV Testing with Diffusion-Driven Network Samples
Researchers have developed a new framework called Policy-Embedded Graph Expansion (PEGE) to improve the efficiency of HIV testing. This approach embeds a generative distribution over graph expansions directly into the decision-making policy, rather than attempting to reconstruct the network topology. Complementing PEGE is Dynamics-Driven Branching (DDB), a diffusion-based graph expansion model designed for data-limited scenarios, which supports decision-making within PEGE. Experiments on real HIV transmission networks demonstrated that the combined PEGE and DDB approach significantly outperformed existing methods, achieving a 17.3% improvement in discounted reward and detecting 15.4% more HIV cases while testing only 25% of the population. AI
IMPACT This AI-driven approach could significantly improve public health outcomes by making disease testing more efficient and targeted.