Competitive algorithms for the weighted list update problem

In this paper we present some deterministic and randomized algorithms for the Weight List Update Problem. In this framework a cost (weight) is associated to each item. The algorithms consist in modifying the well known Move-To-Front heuristic by adding randomness or counters in order to decide whether moving the accessed item. We prove that Random Move-To-Front and Counting Move-To-Front are 2-competitive against any static adversary, and that deterministic Move-To-Front does not share this property. We apply this approach to the management of non-modifiable trees by means of lists of successors proving that 2-competitivity property still holds.