“On the Ability of the Optimal Perceptron to Generalize”, M. Opper, W. Kinzel, J. Kleinz, R. Nehl1990 (MLP NN)⁠:

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. [cf. “double descent”]