Transfer Learning in High-dimensional Ising Models
2026-07-03 • Machine Learning
Machine Learning
AI summaryⓘ
The authors propose a method called Trans-Ising to better estimate models when data is limited and comes from multiple sources with unknown relevance. Their method first picks which extra data sources are helpful by testing them on a small part of the main data to avoid hurting the model. Then, it combines all chosen data to get an initial estimate and finally fine-tunes it using only the main data. The authors show mathematically that their method is reliable and that it works better than just using the main data or naively combining all data. They also confirm these findings through simulations and real-world data.
Ising modelhigh-dimensional estimationtransfer learningpseudolikelihoodnodewise logistic regressionregularizationfolded-concave penaltygraph selection consistencynegative transfer
Authors
Joonho Kim, Seyoung Park
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 $\ell_1$-regularized logistic regression, followed by a target-only correction step using a folded-concave penalty. Theoretically, we establish fixed-node $\ell_2$ and $\ell_1$ 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.