Geometric Amortization of Enumeration Algorithms
share
In this paper, we introduce the technique of geometric amortization for enumeration algorithms. This technique can be used to make the delay of enumeration algorithms more regular without much overhead on the space it uses. More precisely, we are interested in enumeration algorithms having incremental linear delay, that is, algorithms enumerating a set A of size K such that for every t ≤ K, it outputs at least t solutions in time O(tp), where p is the incremental delay of the algorithm. While it is folklore that one can transform such an algorithm into an algorithm with delay O(p), the naive transformation may blow the space exponentially. We show that, using geometric amortization, such an algorithm can be transformed into an algorithm with delay O(plog K) and O(slog K) space, where s is the space used by the original algorithm. We apply geometric amortization to show that one can trade the delay of flashlight search algorithms for their average delay modulo a factor of O(log K). We illustrate how this tradeoff may be advantageous for the enumeration of solutions of DNF formulas.
READ FULL TEXT