Several classes of graph optimization problems, which can be solved us
ing dynamic programming, are known to have more efficient tailor-made
algorithms. This paper discusses four such classes and the underlying
constraints on their subproblem interrelationships that yield these ef
ficient algorithms. These classes are also extended to handle more gen
eral cost functions.