This myth is widespread. It creates the mistaken impression that the process of
decomposing the original problem into subproblems necessarily yields smaller
and smaller problems, until ultimately the subproblems become trivial. By
implication this means that the solution procedure starts with “ . . . the smallest,
and hence simplest, subinstances . . . ” (Brassard and Bartley, 1988, p. 142).
This common view of DP completely distorts the fundamental idea of DP, which
does not rely at all on the subproblems becoming “ smaller ” or “ simpler ”. For
example, in the case of shortest path problems with cycles, the subproblems do
not become smaller nor simpler than the original problem. And in the case of
infinite horizon problems, the subproblems are neither smaller nor easier than
the original problem.