Boosting learning in the presence of incomplete data

Abstract
Boosting techniques have attracted increasing attention in both machine learning and statistical research. While various methods have been developed for different settings, most are designed primarily for complete datasets, limiting their applicability to handle incomplete data such as missing observations and censored data. To address the challenges posed by incomplete data, we introduce boosting estimation techniques tailored specifically for such scenarios. By accounting for the missing data effects, we develop an implementation algorithm using a functional gradient descent and evaluate its performance through numerical studies in finite sample settings.