Online Algorithm for Aggregating Experts' Predictions with Unbounded Quadratic Loss

TL;DR

This paper introduces an online algorithm for aggregating expert predictions under quadratic loss without needing prior loss bounds, utilizing exponential reweighting to adaptively combine expert advice.

Contribution

It presents a novel online aggregation algorithm that operates without pre-known loss bounds, advancing adaptive prediction methods.

Findings

Algorithm effectively aggregates predictions with unbounded quadratic loss.

No prior knowledge of loss bounds is required for the algorithm.

The method is based on exponential reweighting of expert losses.

Abstract

We consider the problem of online aggregation of expert predictions with the quadratic loss function. We propose an algorithm for aggregating expert predictions which does not require a prior knowledge of the upper bound on the losses. The algorithm is based on the exponential reweighing of expert losses.