Key points are not available for this paper at this time.
A linearly separable Boolean function is derived from a set of examples by a perceptron with optimal stability. The probability to reconstruct a pattern which is not learnt is calculated analytically using the replica method.
Opper et al. (Thu,) studied this question.