This technique starts from the initial problem set, and divides it into small subproblems. After the solution to a subprogram has been determined, we store the result to that particular subproblem. In the future, when this subproblem is encountered, we only return its pre-computed result.