¹³It should be remarked, though, that any minor change in the implementation of even a “standard” optimization algorithm like Polak-Ribiere can significantly affect the required number of iterations, so one should view these numbers only as rough or qualitative indications of the learning effort involved in modelling. As a general observation, it has been noticed that the required iteration counts normally stay within the same order of magnitude, but it is not uncommon to have variations of a factor two or three due to, for instance, a different random initialization of neural network parameters.