Property testing in high-dimensional Ising models (Q2328049)

The Ising model, introduced in statistical physics for spin systems, is a \(d\)-vector of binary random variables \(X = (X_1,X_2, \dots, X_d)\) having a probability distribution based on a certain graph. Probability distribution models of this type have been applied also to image analysis, bioinformatics and social networks. In the present paper, the inferential problem of property testing is studied, i.e., testing whether the graph obeys certain binary properties, such as connectedness, cycle presence or maximum clique size. Moreover, correlation based tests are designed, and models including ferromagnetic and antiferromagnetic interactions are considered.