On the bichromatic k -set problem
Timothy M. Chan
Abstract
Timothy M. Chan
Abstract
We study a bichromatic version of the well-known k-set problem : given two sets R and B of points of total size n and an integer k , how many subsets of the form (R ∩ h ) ∪ ( B ∖ h ) can have size exactly k over all halfspaces h ? In the dual, the problem is asymptotically equivalent to determining the worst-case combinatorial complexity of the k-level in an arrangement of n halfspaces . Disproving an earlier conjecture by Linhart [1993], we present the first nontrivial upper bound for all k ≪ n in two dimensions: O ( nk 1/3 + n 5/6−ϵ k 2/3+2 ϵ + k 2 ) for any fixed ϵ<0. In three dimensions, we obtain the bound O ( nk 3/2 + n 0.5034 k 2.4932 + k 3 ). Incidentally, this also implies a new upper bound for the original k -set problem in four dimensions: O ( n 2 k 3/2 + n 1.5034 k 2.4932 + n k 3 ), which improves the best previous result for all k ≪ n 0.923 . Extensions to other cases, such as arrangements of disks, are also discussed.
OpenAlex reports 4 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
We study a bichromatic version of the well-known k-set problem : given two sets R and B of points of total size n and an integer k , how many subsets of the form (R ∩ h ) ∪ ( B ∖ h ) can have size exactly k over all halfspaces h ? In the dual, the problem is asymptotically equivalent to determining the worst-case combinatorial complexity of the k-level in an arrangement of n halfspaces . Disproving an earlier conjecture by Linhart [1993], we present the first nontrivial upper bound for all k ≪ n in two dimensions: O ( nk 1/3 + n 5/6−ϵ k 2/3+2 ϵ + k 2 ) for any fixed ϵ<0. In three dimensions, we obtain the bound O ( nk 3/2 + n 0.5034 k 2.4932 + k 3 ). Incidentally, this also implies a new upper bound for the original k -set problem in four dimensions: O ( n 2 k 3/2 + n 1.5034 k 2.4932 + n k 3 ), which improves the best previous result for all k ≪ n 0.923 . Extensions to other cases, such as arrangements of disks, are also discussed.
Key concepts: Combinatorics, Conjecture, Upper and lower bounds, Mathematics, Integer (computer science), Set (abstract data type), Discrete mathematics, Computer science