Online conformal prediction with decaying step sizes
About
We introduce a method for online conformal prediction with decaying step sizes. Like previous methods, ours possesses a retrospective guarantee of coverage for arbitrary sequences. However, unlike previous methods, we can simultaneously estimate a population quantile when it exists. Our theory and experiments indicate substantially improved practical properties: in particular, when the distribution is stable, the coverage is close to the desired level for every time point, not just on average over the observed sequence.
Anastasios N. Angelopoulos, Rina Foygel Barber, Stephen Bates• 2024
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Conformal Prediction | Changepoint simulated (test) | Coverage90 | 24 | |
| Post-shift coverage recovery | Changepoint Shift 1 (Time 1) | Recovery Time6 | 24 | |
| Time-series interval forecasting | Google Stock | Coverage90.7 | 24 | |
| Post-shift coverage recovery | Changepoint Shift 2 (Time 2) | Recovery Time0.00e+0 | 24 | |
| Time-series interval forecasting | Temperature | Coverage90.4 | 24 | |
| Online Conformal Prediction | Heavy-tailed | Coverage90.8 | 24 | |
| Conformal Prediction | Distribution Drift simulated (test) | Coverage90.6 | 24 | |
| Time-series interval forecasting | Electricity Demand | Coverage90.1 | 24 | |
| Online Conformal Prediction | Variance Changepoint | Coverage0.902 | 24 | |
| Time-series interval forecasting | Amazon Stock | Coverage89.9 | 24 |
Showing 10 of 12 rows