YAML Metadata Warning:empty or missing yaml metadata in repo card
Check out the documentation for more information.
Experiment 5 - AutoGluon Physics Simulator
Overview
AutoGluon 1.5 Extreme model that learns ERA5 temperature correction patterns from ~40 ISD stations surrounding RKSI (Incheon Intl Airport). RKSI is held out during training to validate spatial generalization.
Scores
| Metric | Value |
|---|---|
| RKSI Holdout MAE | 0.969C |
| Holdout Bias | -0.596C |
| Holdout RMSE | 1.206C |
| Training Rows | 1,099,070 |
Best Model: WeightedEnsemble_L4 (3-layer stacking, 7 models)
Hyperparameters
- Preset: extreme
- Time limit: 14,400s (4 hours)
- Bag folds: 8, Stack levels: 2
- Refit full: True
- Dynamic stacking: False
- Training: 40 stations x 10 years (2005-2014)
- Hardware: 8x NVIDIA B200, 192 CPUs, 2TB RAM
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support