Fig. 2
The Eighty Five Percent Rule applied to the Perceptron. a The relative precision, β/βmax, as a function of training error rate and training duration. Training at the optimal error rate leads to the fastest learning throughout. b The dynamics of learning agree well with the theory
