On the On-line Number of Snacks Problem

Weimin Ma,Jane You,Yinfeng Xu,James Liu,Kanliang Wang

DOI: https://doi.org/10.1023/A:1021253103047

2002-01-01

Journal of Global Optimization

Abstract:In the number of snacks problem ( NSP ), which was originally proposed by our team, an on-line player is given the task of deciding how many shares of snacks his noshery should prepare each day. The on-line player must make his decision and then finish the preparation before the customers come to his noshery for the snacks; in other words, he must make decision in an on-line fashion. His goal is to minimize the competitive ratio, defined as ∈ σ: C A (σ)/C OPT (σ), where σ denotes a sequence of numbers of customers, C OPT (σ) is the cost of satisfying σ by an optimal off-line algorithm, and C A (σ) is the cost of satisfying σ by an on-line algorithm. In this paper we give a competitive algorithm for on-line number of snacks problem P 1, the Extreme Numbers Harmonic Algorithm ( ENHA ), with competitive ratio 1+pċ(M-m)/(M+m), where M and m are two extreme numbers of customers over the total period of the game, and p is a ratio concerning the cost of the two types of situations, and then prove that this competitive ratio is the best one if an on-line player chooses a fixed number of shares of snacks for any sequence of numbers of customers. We also discuss several variants of the NSP and give some results for it. Finally, we propose a conjecture for the on-line NSP .