English

ShakyPrepend: A Multi-Group Learner with Improved Sample Complexity

Machine Learning 2026-03-10 v1

Abstract

Multi-group learning is a learning task that focuses on controlling predictors' conditional losses over specified subgroups. We propose ShakyPrepend, a method that leverages tools inspired by differential privacy to obtain improved theoretical guarantees over existing approaches. Through numerical experiments, we demonstrate that ShakyPrepend adapts to both group structure and spatial heterogeneity. We provide practical guidance for deploying multi-group learning algorithms in real-world settings.

Keywords

Cite

@article{arxiv.2603.07319,
  title  = {ShakyPrepend: A Multi-Group Learner with Improved Sample Complexity},
  author = {Lujing Zhang and Daniel Hsu and Sivaraman Balakrishnan},
  journal= {arXiv preprint arXiv:2603.07319},
  year   = {2026}
}

Comments

29 pages, 10 figures, submitted to ICML2026

R2 v1 2026-07-01T11:08:40.280Z