A way to choose a model class
We want to get a low error rate on unseen data.
This is called “structural risk minimization”
It would be really helpful if we could get a guarantee of
the following form:
Test error rate =< train error rate + f(N, h, p)
Where N = size of training set,
            h = measure of the model complexity,
            p = the probability that this bound fails
We need p to allow for really unlucky test sets.
Then we could choose the model complexity that
minimizes the bound on the test error rate.