and discuss the fundamental connection of it
with PAC learnability.
This connection is usually established via the
Vapnik-Chervonenkis dimension (VC dimension).
However, in this talk we will abstract away the definition
of the VC dimension and we will show that instead of the VC dimension
one can use any complexity measure
which is defined on sets of boolean functions and possesses
two, rather natural, properties.
The lecture will be self-contained and will take roughly 1 hour.