The authors study distributional online calibration over an arbitrary convex forecast set Y and under an arbitrary error norm.
Harmonic weighting makes online calibration polynomial in forecast dimension
The authors obtain calibration after d^{O(1/ε)} rounds for both simultaneous binary events and multiclass predictions.
Big Tech
Maxwell Fishelson · Mehryar Mohri
Institute for Advanced Study · Google Research · Courant Institute of Mathematical Sciences, New York
Research Digest··2 min read
Fishelson and Mohri give a single online algorithm for calibrating high-dimensional forecasts without assuming a stochastic model for outcomes.
Why this paper
From Google Research and 2 others
In one line
Harmonic weighting achieves ε-calibration in d^{O(1/ε)} rounds for multi-event and multi-class forecasting.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§