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.