High-Dimensional Graphical Model Selection Using $\ell_1$-Regularized Logistic Regression