On Robust Learning of Ising Models

Erik M. Lindgren, Vatsal Shah, Yanyao Shen, Alexandros G. Dimakis, Adam Klivans

Abstract: Ising Models are one of the most popular class of probability distributions withapplications in wide ranging fields such as physics, engineering and finance. Inthis paper, we attempt to learn the underlying graphical model robustly in presenceof adversarial corruptions. In this work, we establish new lower and upper boundsfor robustly learning Ising models.