Heterogeneous K-core percolation on hypergraphs

In complex systems, there are pairwise and multiple interactions among elements, which can be described as hypergraphs. K-core percolation is widely utilized in the investigation of the robustness of systems subject to random or targeted attacks. However, the robustness of nodes usually correlates with their characteristics, such as degree, and exhibits heterogeneity while lacking a theoretical study on the K-core percolation on a hypergraph. To this end, we constructed a hyperedge K-core percolation model that introduces heterogeneity parameters to divide the active hyperedges into two parts, where hyperedges are inactive unless they have a certain number of active nodes. In the stage of pruning process, when the number of active nodes contained in a hyperedge is less than its set value, it will be pruned, which will result in the deletion of other hyperedges and ultimately trigger cascading failures. We studied the magnitude of the giant connected component and the percolation threshold of the model by mapping a random hypergraph to a factor graph. Subsequently, we conducted a large number of simulation experiments, and the theoretical values matched well with the simulated values. The heterogeneity parameters of the proposed model have a significant impact on the magnitude of the giant connected component and the type of phase transition in the network. We found that decreasing the value of heterogeneity parameters renders the network more fragile, while increasing the value of heterogeneity parameters makes it more resilient under random attacks. Meanwhile, as the heterogeneity parameter decreases to 0, it may cause a change in the nature of network phase transition, and the network shows a hybrid transition.