Optimal and nearly optimal weighted skip lists

Abstract

We consider the problem of building a static (i.e. no updates are performed) skip list of n elements, given these n elements and the corresponding access probabilities or weights. We develop a dynamic programming algorithm that builds an optimal skip list in the sense that the average access cost is minimized. We also consider nearly optimal skip lists, whose average access cost is not optimal but good enough, and can be built more efficiently than optimal skip lists. Several related issues are also discussed, for instance, other approaches to the construction of nearly optimal skip lists or the construction of optimal skip lists that minimize different kinds of search costs.