We analyze the distribution of the running time of Mulmuley’s randomized algorithm for computing the intersections of n given segments in the plane. Its expectation has been known to be O(n log n + k), where k is the number of intersections. We show that for values of k not too close to n (k ≥ Cnlog15n for a large enough constant C), the running time is sharply concentrated around the expected value; e.g., the probability that the expected value is exceeded more than twice is O(n-c), where c can be an arbitrarily large constant (its choice determines the value of C needed in the assumption). Our proof uses an isoperimetric inequality for permutations.
CITATION STYLE
Matoušek, J., & Seidel, R. (1992). A tail estimate for Mulmuley’s segment intersection algorithm. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 623 LNCS, pp. 427–438). Springer Verlag. https://doi.org/10.1007/3-540-55719-9_94
Mendeley helps you to discover research relevant for your work.