A faster method for computing Gama-Nguyen-Regev's extreme pruning coefficients

Abstract: This paper considers Gama-Nguyen-Regev's strategy [GNR10] for optimizing pruning coefficients for lattice vector enumeration. We give a table of optimized coefficients and proposes a faster method for computing near-optimized coefficients for any parameters by interpolation. From the first version published in 2014, we inserted new Section 3.3 to introduce our recent technique to compute approximations of enumeration cost and success probability; both are completed in O(n^2) floating point operations where n is the lattice dimension. For readers who are interested in this topic, we keep the descriptions of our heuristic optimization method in Section 4 although they are outdated now.