FPT-Algorithms for the l-Matchoid Problem with Linear and Submodular Objectives

11/12/2020 ∙ by Chien-Chung Huang, et al. ∙ 0

We design a fixed-parameter deterministic algorithm for computing a maximum weight feasible set under a ℓ-matchoid of rank k, parameterized by ℓ and k. Unlike previous work that presumes linear representativity of matroids, we consider the general oracle model. Our result, combined with the lower bounds of Lovasz, and Jensen and Korte, demonstrates a separation between the ℓ-matchoid and the matroid ℓ-parity problems in the setting of fixed-parameter tractability. Our algorithms are obtained by means of kernelization: we construct a small representative set which contains an optimal solution. Such a set gives us much flexibility in adapting to other settings, allowing us to optimize not only a linear function, but also several important submodular functions. It also helps to transform our algorithms into streaming algorithms. In the streaming setting, we show that we can find a feasible solution of value z and the number of elements to be stored in memory depends only on z and ℓ but totally independent of n. This shows that it is possible to circumvent the recent space lower bound of Feldman et al., by parameterizing the solution value. This result, combined with existing lower bounds, also provides a new separation between the space and time complexity of maximizing an arbitrary submodular function and a coverage function in the value oracle model.



There are no comments yet.


page 1

page 2

page 3

page 4

This week in AI

Get the week's most popular data science and artificial intelligence research sent straight to your inbox every Saturday.