Delta-rule memory improves flu forecasts with seasonal adaptivity
TERN: A Delta-rule Memory with a Seasonal Reference and Online Adaptation for Epidemic Forecasting
Machine LearningArtificial Intelligence
Summary
Flu outbreaks change every year and can be tough to predict with limited past data. The authors developed a new forecasting tool called TERN that remembers earlier flu seasons but also adapts quickly to current changes. TERN uses a special memory system that updates when new information comes in and pays attention to different phases of the flu season. Tested on public flu data, TERN made more accurate predictions than other common forecasting methods.
What this means in practice
- •For public health data teams: Improve flu outbreak predictions by integrating a memory that adapts with seasonal changes for better vaccine distribution planning.
- •For weather and climate forecasters: Adapt forecasting models to handle seasonal patterns and phase shifts using a delta-rule fast-weight memory approach inspired by TERN.
Authors
Shunya Nagashima, Yuta Funayama
Abstract
Weekly influenza surveillance counts guide vaccine distribution and public-health alerts, yet they are hard to forecast. Each region offers only a few seasons, waves shift in timing and height every year, and information that helps while a wave grows misleads after its peak, whereas last season's shape stays informative for a year. Existing epidemic graph models and general forecasters read a short fixed window and treat all past information alike, so they neither exploit earlier seasons nor discard stale associations when the epidemic phase changes. To address these limitations, we propose TERN, a forecaster built around a delta-rule fast-weight memory that decays channel-wise and erases along a learned address under gates driven by local epidemic-phase features, combined with an explicit seasonal reference and online adaptation. On three Cola-GNN influenza benchmarks, TERN outperformed epidemic graph models and general forecasters, matched or exceeded seasonal references, and a controlled comparison confirmed the contribution of the memory itself.