Researchers have developed a new method called adaptive cross-bagging to improve the reproducibility of machine learning predictions. This technique addresses the instability caused by random seeds in machine learning algorithms, which can affect downstream estimators. The proposed method formalizes random seed stability and proves that subbagging can ensure stability for bounded-outcome regression algorithms. Numerical experiments demonstrate that adaptive cross-bagging achieves the desired stability with only a minor computational cost compared to other methods. AI
Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →
RANK_REASON The item is an arXiv preprint detailing a new methodology for improving machine learning reproducibility.
