Influenza hospitalization forecasting
View
←
1 · Model progression
2 · Fit the center
3 · Fit the spread
→
From separate models to shared learning
Regularized
VAR
State-specific
LightGBM
Separate fits
Ensembling throughout
Texas
California
New York
Other locations
Pooled
LightGBM
Shared training
across locations
Fit the center
Squared-error loss
Fit the spread
Negative log-likelihood
Time
Center
Spread
forecast quantiles