Inductive approach
Take smaller instances of the
problem.
Solve the smaller instances to
get policies.
Use various ground policies for
smaller instances to learn first
order policy.