Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
4952368 | Theoretical Computer Science | 2017 | 18 Pages |
Abstract
In this paper, we improve upper and lower bounds on the competitive ratio of k-OFTM. Our main result is to improve an upper bound of O(k2) by Kesselman et al. to 5B+âB/kââ4âB/2kâ=O(k) for Bâ¥2k. Note that this upper bound is tight up to a multiplicative constant factor since the lower bound given by Kesselman et al. is Ω(k). We also give two lower bounds. First we give a lower bound of 2BâB/(kâ1)â+1 on the competitive ratio of deterministic online algorithms for any kâ¥2 and any Bâ¥kâ1, which improves the previous lower bound of Bâ2B/kâ by a factor of almost four. Next, we present the first nontrivial lower bound on the competitive ratio of randomized algorithms. Specifically, we give a lower bound of kâ1 against an oblivious adversary for any kâ¥3 and any B. Since a deterministic algorithm, as mentioned above, achieves an upper bound of about 10k, this indicates that randomization does not help too much.
Related Topics
Physical Sciences and Engineering
Computer Science
Computational Theory and Mathematics
Authors
Jun Kawahara, Koji M. Kobayashi, Shuichi Miyazaki,