Read the original document by opening this link in a new tab.
Table of Contents
Abstract
1. Introduction
1.1 Outline
2. Formulation of the Estimation Problem
2.1 Data, Likelihood and the Statistical Model
2.2 Statistical Target Parameter
Summary
In this work, the authors introduce the Personalized Online Super Learner (POSL), an online ensembling algorithm for streaming data that optimizes predictions with respect to baseline covariates, allowing for varying degrees of personalization. POSL leverages a diversity of candidate algorithms, including online algorithms with different training and update times, fixed algorithms, pooled algorithms, and individualized algorithms. The algorithm adapts to changing data-generating environments and provides reliable predictions for time-series data. The authors propose a novel online ensembling algorithm, POSL, that aims to optimize baseline-covariate-level forecasts. They discuss the challenges of applying online learning strategies and address the issue of catastrophic forgetting. The article provides theoretical foundations for the proposed algorithm and demonstrates its performance in various simulation studies and a data analysis example for blood pressure forecasting.