Core Concepts
The contention resolution framework is a versatile rounding technique used to solve constrained submodular function maximization problems. This work applies this framework to the hypergraph matching, knapsack, and k-column sparse packing problems, providing non-constructive lower bounds on the correlation gap and constructing monotone contention resolution schemes.
Abstract
The content discusses the contention resolution framework, which is a versatile rounding technique used to solve constrained submodular function maximization problems. The author applies this framework to three specific problems: the hypergraph matching problem, the knapsack problem, and the k-column sparse packing integer problem (k-CS-PIP).
For the hypergraph matching problem, the author adapts techniques from prior work to non-constructively prove that the correlation gap is at least 1-e^(-k)/k and provides a monotone (1-e^(-bk))/bk-balanced contention resolution scheme, generalizing previous results.
For the knapsack problem, the author proves that the correlation gap of instances where exactly k copies of each item fit into the knapsack is at least 1-e^(-2)/2. The author also provides several monotone contention resolution schemes: a (1-e^(-2))/2-balanced scheme for instances where all item sizes are strictly bigger than 1/2, a 4/9-balanced scheme for instances where all item sizes are at most 1/2, and a 0.279-balanced scheme for instances with arbitrary item sizes.
For k-CS-PIP, the author slightly modifies an existing algorithm to obtain a 1/(4k+o(k))-balanced contention resolution scheme and hence a (4k+o(k))-approximation algorithm for k-CS-PIP based on the natural LP relaxation.
The appendix contains technical statements used in the main sections, as well as examples attaining the integrality gap for the three settings considered.
Stats
The author provides the following key metrics and figures:
The correlation gap of the fractional hypergraph matching polytope is at least (1-e^(-k))/k.
The correlation gap of class-k knapsack instances (where exactly k copies of each item fit into the knapsack) is at least (1-e^(-2))/2.
The author's monotone contention resolution schemes achieve the following balancedness:
Hypergraph matching: (1-e^(-bk))/bk
Knapsack (big items): (1-e^(-2))/2
Knapsack (small items): 4/9
Knapsack (general): 0.279
For k-CS-PIP, the author's scheme achieves a balancedness of 1/(4k+o(k)).