no code implementations • 22 Feb 2024 • Baptiste Abélès, Joseph de Vilmarest, Olivier Wintemberger
In this paper, we propose a new way of estimating variances based on online learning theory; we adapt expert aggregation methods to learn the variances over time.