Group Activity Selection with Few Agent Types

TL;DR

This paper explores the computational complexity of the Group Activity Selection Problem (GASP) and its variants when parameterized by the number of agent types, providing new algorithms and complexity bounds.

Contribution

It establishes the complexity map for GASP variants with respect to agent types, introducing novel algorithms and proving W[1]-hardness results using advanced combinatorial techniques.

Findings

Developed a fixed-parameter algorithm for GASP variants.

Proved W[1]-hardness of sGASP with respect to agent types.

Introduced new Subset Sum techniques and compression methods.

Abstract

The Group Activity Selection Problem (GASP) models situations where a group of agents needs to be distributed to a set of activities while taking into account preferences of the agents w.r.t. individual activities and activity sizes. The problem, along with its two previously proposed variants sGASP and gGASP, has been studied in the parameterized complexity setting with various parameterizations, such as number of agents, number of activities and solution size. However, the complexity of the problem parameterized by the number of types of agents, a parameter motivated and proposed already in the paper that introduced GASP, has so far remained open. In this paper we establish the complexity map for GASP, sGASP and gGASP when the number of types of agents is the parameter. Our positive results, consisting of one fixed-parameter algorithm and one XP algorithm, rely on a combination of novel Subset Sum machinery (which may be of general interest) and identifying certain compression steps which allow us to focus on solutions which are "acyclic". These algorithms are complemented by matching lower bounds, which among others answer an open question of Gupta, Roy, Saurabh and Zehavi (2017). In this direction, the techniques used to establish W[1]-hardness of sGASP are of particular interest: as an intermediate step, we use Sidon sequences to show the W[1]-hardness of a highly restricted variant of multi-dimensional Subset Sum, which may find applications in other settings as well.