On the Minimax Regret of Sequential Probability Assignment via Square-Root Entropy

Zeyu Jia, Yury Polyanskiy, Alexander Rakhlin · 2025

Read On the Minimax Regret of Sequential Probability Assignment via Square-Root Entropy by Zeyu Jia, Yury Polyanskiy, Alexander Rakhlin (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-17823.txt.

Description

Zeyu Jia, Yury Polyanskiy, and Alexander Rakhlin study, in 2025, how well probabilities can be assigned step by step as outcomes arrive. The cost of each guess is measured by logarithmic loss, and the benchmark is minimax regret: the gap between the best sequential strategy and the best hindsight choice. They tie that gap to geometric measures of the model class.

Without side information, regret is bounded by a new entropy notion built on the square-root scale, close to the Hellinger distance. With side information, matching upper and lower bounds emerge for well-behaved classes. The results give forecasters a geometric yardstick for sequential prediction.

Catalog record

CREATORZeyu Jia, Yury Polyanskiy, Alexander Rakhlin
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE192 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-17823

Browse the full library · About the Tower of Babel