Ensemble approach for generalized network dismantling

TL;DR

This paper introduces a spectral approximation method using ensemble techniques to address the NP-hard network dismantling problem, aiming to efficiently identify minimal node sets for network fragmentation.

Contribution

It proposes a novel spectral approximation approach combined with ensemble methods to explore the solution landscape of the generalized network dismantling problem.

Findings

Spectral approximation effectively identifies critical nodes for network dismantling.

Ensemble methods reveal diverse solutions and improve robustness.

The approach offers a scalable alternative to exact algorithms for NP-hard problems.

Abstract

Finding a set of nodes in a network, whose removal fragments the network below some target size at minimal cost is called network dismantling problem and it belongs to the NP-hard computational class. In this paper, we explore the (generalized) network dismantling problem by exploring the spectral approximation with the variant of the power-iteration method. In particular, we explore the network dismantling solution landscape by creating the ensemble of possible solutions from different initial conditions and a different number of iterations of the spectral approximation.