Orthogonal Vectors Conjecture

From papers

For positive integers nn and mm, let OV(n,m)\emph{\textsf{OV}}(n,m) denote the Orthogonal Vectors problem: given sets A={x(1),,x(n)}A=\{x^{(1)},\ldots,x^{(n)}\} and B={y(1),,y(n)}{0,1}mB=\{y^{(1)},\ldots,y^{(n)}\}\subset\{0,1\}^m, determine whether there are i,j[n]i,j\in[n] with x(i)y(j)=0\langle x^{(i)}-y^{(j)}\rangle=0. Orthogonal Vectors Conjecture. For every δ>0\delta>0, there is c1c\ge 1 such that OV(n,m)\emph{\textsf{OV}}(n,m) cannot be solved in O(n2δ)O(n^{2-\delta}) time on instances with m=clognm=c\log n. This conjecture is a standard source of conditional fine-grained lower bounds, including reductions to kernel density estimation; its resolution status is not specified in the input.

Progress summary

Nothing recorded yet. Refresh searches the literature and the public web for attempts on this problem, and writes the first summary here.

Sources & referencesView supporting material

Primary source

Josh Alman and Yunfeng Guan, “Finer-Grained Hardness of Kernel Density Estimation”, arXiv:2407.02372 (2024).

Solutions 0

No solutions have been posted yet.