English

Transfer Learning in High-dimensional Ising Models

Machine Learning 2026-07-03 v1 Methodology Machine Learning

Abstract

In high-dimensional Ising model estimation, target sample sizes are often limited, and effectively using auxiliary binary datasets of unknown relevance remains challenging. To address this, we propose Trans-Ising, a transfer learning method that combines a loss-based source screening rule with a two-stage estimation procedure. The method first identifies informative auxiliary sources using held-out target pseudolikelihood to prevent negative transfer. It then computes an initial estimator via pooled nodewise 1\ell_1-regularized logistic regression, followed by a target-only correction step using a folded-concave penalty. Theoretically, we establish fixed-node 2\ell_2 and 1\ell_1 error bounds, exact graph selection consistency, and the conditional consistency of the screening rule. Through extensive simulations and real-data analyses, we demonstrate that Trans-Ising achieves lower estimation errors than both target-only estimation and naive data pooling.

Cite

@article{arxiv.2607.03005,
  title  = {Transfer Learning in High-dimensional Ising Models},
  author = {Joonho Kim and Seyoung Park},
  journal= {arXiv preprint arXiv:2607.03005},
  year   = {2026}
}

Comments

Published as a conference paper at ICML 2026