From the source
Google Research is presenting five partner-driven case studies showing how its Population Dynamics Foundation Model (PDFM) can plug into existing epidemiological workflows to improve predictions across diseases and regions.
- PDFM compresses privacy-preserving search trends, mobility, built-environment density and environmental data into location embeddings refreshed monthly, without requiring task-specific fine-tuning.
- In a US-Canada border study, PDFM embeddings explained 36% more variance in MMR vaccination coverage than conventional inputs by capturing cross-border mobility.
- The model also showed statistically significant gains in dengue forecast accuracy in Mexico and cholera outbreak prediction in the Democratic Republic of the Congo.
