2026-10-12T15:46:41Z https://api.figshare.com/v2/oai

oai:figshare.com:article/34035836 2026-09-30T17:41:49Z category_1 category_15 category_21 category_734 category_931 category_132 portal_5 item_type_6 month_year_09_2026

<p>Supplementary methods and results.</p> Joseph Lemaitre (25145381) Justin Lessler (116180) Biophysics Neuroscience Biotechnology Biological Sciences not elsewhere classified Information Systems not elsewhere classified Infectious Diseases somewhat overconfident projections leading ensemble methods cdc flusight challenges capture multimodal uncertainty achieves forecast accuracy 2024 &# 8211 2022 &# 8211 influpaint generates realistic encoding influenza seasons provide important information generative diffusion models diverse epidemic trajectories diffusion models influpaint learns influenza dynamics simulated trajectories epidemic dynamics time evaluation spatiotemporal images rich distribution retrospective evaluation partial observations hybrid dataset highly accurate flexible framework emergent trends disease dynamics current mechanistic conditional generation best performance <div><p>Forecasting infectious disease incidence can provide important information to guide public health planning, yet it is difficult because epidemic dynamics are complex. Current mechanistic and statistical approaches often struggle to capture multimodal uncertainty or emergent trends. Influpaint adapts denoising diffusion probabilistic models to epidemic forecasting. By encoding influenza seasons as spatiotemporal images in which pixel intensity represents incidence, Influpaint learns a rich distribution of disease dynamics from a hybrid dataset of surveillance and simulated trajectories. Forecasting is formulated as a conditional generation (inpainting) task from partial observations. We show that Influpaint generates realistic, diverse epidemic trajectories and achieves forecast accuracy that is competitive with leading ensemble methods in retrospective evaluation. In real-time evaluation during the 2022–2025 U.S. CDC FluSight challenges, performance improved substantially across seasons, with highly accurate but somewhat overconfident projections in 2024–2025. The best performance was achieved with a training dataset containing 30% surveillance and 70% simulated trajectories. These results show that diffusion models can capture important spatiotemporal structure in influenza dynamics and provide a flexible framework for probabilistic infectious disease forecasting.</p></div> 2026-09-30T18:00:00Z Text Journal contribution 10.1371/journal.pcbi.1014846.s001 https://figshare.com/articles/journal_contribution/_p_Supplementary_methods_and_results_p_/34035836 CC BY 4.0