Abstract Given n real numbers, the e-CLOSENESS problem consists in deciding whether any two of them are within e of each other, where e is a fixed parameter of the problem. This is a basic problems with an Ω ( n log n ) lower bound on the running time in the algebraic computation tree model. In this paper we show that for a natural approximation version thereof the same lower bound holds. The main tool used is a lower bound theorem of Ben-Or. We introduce a new interpolation method relating the approximation version of the problem to two corresponding exact versions. Using this result, we are able to prove the optimality of the running time of a cartographic algorithm, that determines an approximate solution to the so-called MAP LABELING problem. MAP LABELING was shown to be NP-complete. The approximation algorithm discussed here is of provably optimal approximation quality. Our result is the first n log n lower bound for an approximate problem. The proof method is general enough to be potentially helpful for further results of this type.
[1]
E. Imhof.
Positioning Names on Maps
,
1975
.
[2]
Donald E. Knuth,et al.
The Problem of Compatible Representatives
,
1992,
SIAM J. Discret. Math..
[3]
L FredmanMichael,et al.
On the complexity of computing the measure of ∪[ai,bi]
,
1978
.
[4]
Michael Ben-Or,et al.
Lower bounds for algebraic computation trees
,
1983,
STOC.
[5]
Bruce W. Weide,et al.
On the complexity of computing the measure of ∪[ai,bi]
,
1978,
CACM.
[6]
Frank Wagner,et al.
A packing problem with applications to lettering of maps
,
1991,
SCG '91.