mlsanityData to 7 Oct 2026

An Elementary Predictor Obtaining 2√T+1 Distance to Calibration

arXiv 2402.11410 ↗First posted 18 Feb 2024cs.LG cs.DS stat.MLcited by 1

Not on the map. Why?

Abstract (arXiv)

Blasiok et al. [2023] proposed distance to calibration as a natural measure of calibration error that unlike expected calibration error (ECE) is continuous. Recently, Qiao and Zheng [2024] gave a non-constructive argument establishing the existence of an online predictor that can obtain O(√T) distance to calibration in the adversarial setting, which is known to be impossible for ECE. They leave as an open problem finding an explicit, efficient algorithm. We resolve this problem and give an extremely simple, efficient, deterministic algorithm that obtains distance to calibration error at most 2√T+1.

Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Builds on 5 of its 8 references that are in the map. First posted on arXiv 18 Feb 2024.

Builds on

5 of 8 references.

References · 5

  1. 2022A Unifying Theory of Distance from Calibrationcited by 51
  2. 2020Stronger Calibration Lower Bounds via Sidesteppingcited by 38
  3. 2022Smooth Calibration, Leaky Forecasts, Finite Recall, and Nash Dynamicscited by 22
  4. 2024On the Distance from Calibration in Sequential Predictioncited by 18
  5. 2022Faster online calibration without randomization: interval forecasts and the power of two choicescited by 4

Built on by

Cited by year

Citing papers per year: 2024: 1.

1 paper.