The question is how to make a network that is both "economical" and "robust" without taking up too much space. This is a theoretical question worked on by the great Russian mathematician Andrey Kolmogorov (1903-1987). But this theoretical question also conditions the way in which we can build a computer network or... a human brain: to be intelligent without having a big head, you need a neural network that is efficient BUT ALSO compact! Mention Szemeredi lemma?