An Active Learning Method Based on Bagging and Applied to Surrogate Modelling
Nomena Andrianarisoa, Matthieu Ancellin, Vincent Laurent
⚠ This is a preprint. It may change before it is accepted for publication.

Abstract

In this work, we study active learning techniques to highlight their strengths and limits in the context of surrogate modeling. We implemented an iterative algorithm that uses active sampling methods to approximate known functions. Through a series of experiments, we demonstrate the effectiveness of active learning in improving surrogate models by wisely selecting informative data points. Furthermore, we examine cases where active learning may face challenges, particularly when balancing exploration and exploitation. This study provides information on the improvement of the efficiency and accuracy of surrogate models in practical applications.

Download