A Unified Approach to Learning Ising Models: Beyond Independence and Bounded Width

Abstract

We revisit the well-studied problem of e ciently learning the underlying structure and parameters of an Ising model from data. Current algorithmic approaches achieve essentially optimal sample complexity when samples are generated i.i.d. from the stationary measure and the underlying model satis es “width” constraints that bound the total ℓ1 interaction involving each node. However, these assumptions are not satis ed in some important settings of interest, like temporally correlated data or more complicated models (like spin glasses) that do not satisfy width bounds. We analyze a simple existing approach based on node-wise logistic regression, and show it provably succeeds at e ciently recovering the underlying Ising model in several new settings: (1) Given dynamically generated data from a wide variety of Markov chains, including Glauber, block, and round-robin dynamics, logistic regression recovers the parameters with sample complexity that is optimal up to log log factors. This generalizes the specialized algorithm of Bresler, Gamarnik, and Shah (IEEE Trans. Inf. Theory ’18) for structure recovery in bounded degree graphs from Glauber dynamics. (2) For the Sherrington-Kirkpatrick model of spin glasses, given poly() independent samples, logistic regression recovers the parameters in most of the proven high-temperature regime via a simple reduction to weaker structural properties of the measure. This improves on recent work of Anari, Jain, Koehler, Pham, and Vuong (SODA ’24) which gives distribution learning at higher temperature. (3) As a simple byproduct of our techniques, logistic regression achieves an exponential improvement in learning from samples in the M-regime of data considered by Dutt, Lokhov, Vu ray, and Misra (ICML ’21) as well as novel guarantees for learning from the adversarial Glauber dynamics of Chin, Moitra, Mossel, and Sandon. Our approach thus provides a signi cant generalization of the elegant analysis of logistic regression by Wu, Sanghavi, and Dimakis (Neurips ’19) without any algorithmic modi cation in each setting