Evolutionary graph-based v+e optimization for protection against epidemics

Abstract

Protection against spreading threats in networks gives rise to a variety of interesting optimization problems. Among others, vertex protection problems such as the Firefighter Problem and vaccination optimization problem can be tackled. Interestingly, in some cases a networked system can be made more resilient to threats, by changing its connectivity, which motivates the study of another type of optimization problems focused on adapting graph connectivity. In this paper the above-mentioned approaches are combined, that is both vertex and edge protection is applied in order to stop the threat from spreading. Solutions to the proposed problem are evaluated using different cost functions for protected vertices and edges, motivated by real-life observations regarding the costs of epidemics control. Instead of making decisions for each of the vertices and edges a decision model is used (based on rules or a neural network) with parameters optimized using an evolutionary algorithm. In the experiments the model using rules was found to perform better than the one based on a neural network.