Ultrametricity, Hopfield Model and all that

Abstract

It is part of our every day experience that when we try to memorize new information we look for all possible relationships with previously stored patterns. If we can classify the new pattern, that is, place it in a hierarchical tree of categories we do it with so much eagerness that sometimes we just censor the data so as to eliminate some exceptional features. Hopfield’s model⁽¹⁾ for the brain seems, on the contrary, to be quite happy crunching uncorrelated orthogonal words. This is the more surprising given the fact that otherwise the model simulates quite nicely other features of Human Memory.