Skip to content

Memory-hedged Exponentially Weighted Least-Squares Online aggregation (MELO)

Home / Concepts / Memory-hedged Exponentially Weighted Least-Squares Online aggregation (MELO)

Memory-hedged Exponentially Weighted Least-Squares Online aggregation (MELO)

Auto-generated stub. Edit this file to add more details.

A model-agnostic online aggregation framework that hedges across multiple adaptation scales using EWLS-adapted forecasts to maintain stability and adapt to distribution shifts.

Why It Matters

MELO introduces a novel model-agnostic framework for online prediction under distribution shift by dynamically aggregating forecasts across multiple forgetting factors without needing external regime markers.

Evidence

We propose MELO (Memory-hedged Exponentially Weighted Least-Squares Online aggregation), a model-agnostic method that hedges across adaptation scales

Metadata & Links