Ginestra Bianconi,Sergey N. Dorogovtsev
Abstract:Hypergraphs are higher-order networks that capture the interactions between two or more nodes. Hypergraphs can always be represented by factor graphs, i.e. bipartite networks between nodes and factor nodes (representing groups of nodes). Despite this universal representation, here we reveal that $k$-core percolation on hypergraphs can be significantly distinct from $k$-core percolation on factor graphs. We formulate the theory of hypergraph $k$-core percolation based on the assumption that a hyperedge can only be intact if all its nodes are intact. This scenario applies for instance to supply chains where the production of a product requires all raw materials and all processing steps; in biology it applies to protein-interaction networks where protein complexes can only function if all the proteins are present, and it applies as well to chemical reaction networks where a chemical reaction can take place only when all the reactants are present. Formulating a message-passing theory for hypergraph $k$-core percolation, and combining it with the theory of critical phenomena on networks we demonstrate sharp differences with previously studied factor graph $k$-core percolation processes where it is allowed for hyperedges to have one or more damaged nodes and still be intact. To solve the dichotomy between $k$-core percolation on hypergraphs and on factor graphs, we define a set of pruning processes that act either exclusively on nodes or exclusively on hyperedges and depend of their second-neighborhood connectivity. We show that the resulting second-neighbor $k$-core percolation problems are significantly distinct from each other. Moreover we reveal that although these processes remain distinct from factor graphs $k$-core processes, when the pruning process acts exclusively on hyperedges the phase diagram is reduced to the one of factor graph $k$-cores.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is the significant difference between the **hypergraph k - core percolation problem** and the **factor graph k - core percolation problem**. Specifically, the author focuses on the fact that in a hypergraph, when a node is removed or damaged, all the hyper - edges to which it belongs will also be destroyed, which is different from the situation in a factor graph. In a factor graph, even if a certain node is removed, as long as other nodes are still intact, the corresponding hyper - edge can still exist.
### Background of the Paper
A hypergraph is a high - order network that can capture the interactions between two or more nodes. A hypergraph can be represented as a factor graph, that is, a bipartite network between nodes and factor nodes (representing a group of nodes). However, although this representation method is general, the author finds that there are significant differences between k - core percolation on hypergraphs and k - core percolation on factor graphs.
### Main Problems
1. **Definition of Hypergraph k - Core Percolation**:
- A hyper - edge in a hypergraph can remain intact only when all of its nodes are intact. For example, in a supply chain, all raw materials and processing steps are required to produce a product; in biology, a protein complex can function only when all proteins are present; in a chemical reaction network, a chemical reaction can occur only when all reactants are present.
2. **Message - Passing Theory**:
- The author proposes a message - passing theory for hypergraph k - core percolation and, combined with the theory of critical phenomena on networks, shows significant differences from the previously studied factor graph k - core percolation process. In the factor graph k - core percolation process, hyper - edges are allowed to remain intact when some nodes are damaged.
3. **Pruning Processes**:
- To solve the differences between hypergraph k - core percolation and factor graph k - core percolation, the author defines a set of pruning processes that are either node - only or hyper - edge - only and depend on their second - order neighborhood connectivity. The results show that the second - order neighborhood k - core percolation problems generated by these pruning processes are significantly different from each other. Moreover, when the pruning process is hyper - edge - only, the phase diagram is simplified to the phase diagram of the factor graph k - core.
### Research Methods
- **Message - Passing Equations**:
- The author derives the message - passing equations for hypergraph (k, n)-core percolation and, combined with the theory of generating functions, studies the (k, n)-core percolation problem on random hypergraphs. These equations are applicable to any locally tree - like hypergraph and can be verified by simulation.
- **Phase Transitions and Phase Diagrams**:
- The author explores the phase transitions of (k, n)-core percolation on random hypergraphs and obtains phase diagrams. These phase diagrams are more abundant than the k - core percolation phase diagrams on ordinary graphs and factor graphs. In particular, in (2, 2)-core percolation, the phase transition on the factor graph is continuous, while on the hypergraph, there are continuous phase transition lines and mixed phase transition lines, and these two lines converge at the tricritical point.
### Conclusions
By introducing a new class of k - core problems, the author solves the differences between hypergraph k - core percolation and factor graph k - core percolation. These new models not only consider the nearest neighborhood of nodes but also the influence of the second - order neighborhood. The research shows that these models are related to both hypergraph (k, n)-core and factor graph (k, n)-core, but they show different characteristics in specific cases.