Detecting all regular polygons in a point set

Abstract: In this paper, we analyze the time complexity of finding regular polygons in a set of n points. We combine two different approaches to find regular polygons, depending on their number of edges. Our result depends on the parameter alpha, which has been used to bound the maximum number of isosceles triangles that can be formed by n points. This bound has been expressed as O(n^{2+2alpha+epsilon}), and the current best value for alpha is ~0.068. Our algorithm finds polygons with O(n^alpha) edges by sweeping a line through the set of points, while larger polygons are found by random sampling. We can find all regular polygons with high probability in O(n^{2+alpha+epsilon}) expected time for every positive epsilon. This compares well to the O(n^{2+2alpha+epsilon}) deterministic algorithm of Brass.