Extending the Accommodating Function.

Joan Boyar, Lene M. Favrholdt, Kim S. Larsen, and Morten N. Nielsen.
Acta Informatica, 40(1): 3-35, 2003.

The applicability of the accommodating function, a relatively new measure for the quality of on-line algorithms, is extended.

The standard quality measure for on-line algorithms is the competitive ratio, which is, roughly speaking, the worst case ratio of the on-line performance to the optimal off-line performance. However, for many on-line problems, the competitive ratio gives overly pessimistic results and/or fails to distinguish between algorithms that are known to perform very differently in practice. Many researchers have proposed variations on the competitive ratio to obtain more realistic results. These variations are often tailor-made for specific on-line problems.

The concept of the accommodating function applies to any on-line problem with some limited resource, such as bins, seats in a train, or pages in a cache. If a limited amount n of some resource is available, the accommodating function A(alpha) is the competitive ratio when input sequences are restricted to those for which the amount alpha n of resources suffices for an optimal off-line algorithm. For all resource bounded problems, the standard competitive ratio is the limit of A(alpha) for alpha going towards infinity.

The accommodating function was originally used only for alpha>=1. We focus on alpha<1, observe that the function now appears interesting for a greater variety of problems, and use it to make new distinctions between known algorithms and to find new ones.

---

publication

Link to the publication at the publisher's site - subscription may be required.
Text required by the publisher (if any): The final publication is available at link.springer.com.

open access (323 KB)

The same as the publisher's version, when the publisher permits. Otherwise, the author's last version before the publisher's copyright; this is often exactly the same, but sometimes fonts, page numbers, figure numbers, etc. are different. It may also be a full version. However, it is safe to read this version, and at the same time cite the official version, as long as references to concrete locations, numbered theorems, etc. inside the article are avoided.

other publications

Other publications by the author.