A New Algorithm for the Closest Pair of Points for Very Large Data Sets Using Exponent Bucketing and Windowing.

In this contribution, a simple and efficient algorithm for the closest-pair problem in $$E^1$$ is described using the preprocessing based on exponent bucketing and exponent windowing respecting accuracy of the floating point representation. The preprocessing is of the O(N) complexity. Experiments made for the uniform distribution proved significant speedup. The proposed approach is applicable for the $$E^2$$ case.