Prediction with Expert Advice under Discounted Loss

TL;DR

This paper extends the prediction with expert advice framework to scenarios with discounted losses, introducing new algorithms and theoretical bounds for this setting.

Contribution

It generalizes existing algorithms to handle discounted loss accumulation and proposes a new exponential weights variant with proven loss bounds.

Findings

Generalized the Aggregating Algorithm for discounted losses

Developed a new exponential weights algorithm variant

Proved theoretical loss bounds for the proposed algorithms

Abstract

We study prediction with expert advice in the setting where the losses are accumulated with some discounting---the impact of old losses may gradually vanish. We generalize the Aggregating Algorithm and the Aggregating Algorithm for Regression to this case, propose a suitable new variant of exponential weights algorithm, and prove respective loss bounds.