Vapnik-Chervonenkis dimension

Suppose we are in the modelling framework with feature space $A$, domain $B$, and we have hypothesis space $H$. The VC-dimension of $H$ is the size of the largest set $S \subset A$ such that for all $l: S \rightarrow B$ there exists $h_l \in H$ such that $h_l(s) = l(s)$ for all $s \in S$.