FPT Approximation Schemes for Maximizing Submodular Functions
Piotr Skowron
Abstract
Open-access reader
Piotr Skowron
Abstract
Open-access reader
We investigate the existence of approximation algorithms for maximization of submodular functions, that run in fixed parameter tractable (FPT) time. Given a non-decreasing submodular set function $v: 2^X \to \mathbb{R}$ the goal is to select a subset $S$ of $K$ elements from $X$ such that $v(S)$ is maximized. We identify three properties of set functions, referred to as $p$-separability properties, and we argue that many real-life problems can be expressed as maximization of submodular, $p$-separable functions, with low values of the parameter $p$. We present FPT approximation schemes for the minimization and maximization variants of the problem, for several parameters that depend on characteristics of the optimized set function, such as $p$ and $K$. We confirm that our algorithms are applicable to a broad class of problems, in particular to problems from computational social choice, such as item selection or winner determination under several multiwinner election systems.
A significance statement is not available in the OpenAlex record.
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 investigate the existence of approximation algorithms for maximization of submodular functions, that run in fixed parameter tractable (FPT) time. Given a non-decreasing submodular set function $v: 2^X \to \mathbb{R}$ the goal is to select a subset $S$ of $K$ elements from $X$ such that $v(S)$ is maximized. We identify three properties of set functions, referred to as $p$-separability properties, and we argue that many real-life problems can be expressed as maximization of submodular, $p$-separable functions, with low values of the parameter $p$. We present FPT approximation schemes for the minimization and maximization variants of the problem, for several parameters that depend on characteristics of the optimized set function, such as $p$ and $K$. We confirm that our algorithms are applicable to a broad class of problems, in particular to problems from computational social choice, such as item selection or winner determination under several multiwinner election systems.
Key concepts: Submodular set function, Maximization, Separable space, Set (abstract data type), Function (biology), Mathematics, Mathematical optimization, Class (philosophy)