The best student of an Ising perceptron

We suggest a set of questions the student of an Ising perceptron can ask in order to achieve perfect generalization at the lower bound known from information theory. We also show that before that point is reached the generalization performance of this set combined with the Bayes strategy is optimal