Trees = Networks ?!?

This work addresses the intrinsic relationship between trees and networks (i.e. graphs). A complete (invertible) mapping is presented which allows trees to be mapped into weighted graphs and then backmapped into the original tree without loss of information. The extension of this methodology to more general networks, including unweighted structures, is also discussed and illustrated. It is shown that the identified duality between trees and graphs underlies several key concepts and issues of current interest in complex networks, including comprehensive characterization of trees and community detection. For instance, additional information about tree structures (e.g. phylogenetic trees) can be immediately obtained by taking into account several off-the-shelf network measurements – such as the clustering coefficient, degree correlations and betweenness centrality. At the same time, the hierarchical structure of networks, including the respective communities, becomes clear when the network is represented in terms of the respective tree. Indeed, the network-tree mapping described in this work provides a simple and yet effective means of community detection.

💡 Research Summary

The paper “Trees = Networks ?!?” by L. da F. Costa and F. A. Rodrigues investigates a deep, bidirectional relationship between hierarchical tree structures and general graphs. The authors introduce a complete, invertible mapping that converts any rooted tree into a weighted network (called a Tree‑Ancestor Network, TAN) and then reconstructs the original tree without any loss of information. The forward mapping proceeds level by level: nodes belonging to the same branch at a given depth are fully connected with unit weight, while connections between nodes of adjacent levels receive weights proportional to the level difference (1, 2, … 2H, where H is the tree height). This weighting scheme faithfully encodes the tree’s intrinsic distances, ensuring that the resulting weighted graph preserves the exact hierarchical relationships.

The inverse mapping (network‑to‑tree) is based on an agglomerative procedure. The algorithm repeatedly identifies the weakest (smallest‑weight) edge, merges the two incident vertex groups into a single super‑node, and updates the weight matrix by assigning to the new node the average weight of the merged groups. By iterating this process until a single node remains, the original tree topology is recovered exactly. The authors prove mathematically that this back‑mapping is lossless and demonstrate it empirically by showing that adjacency matrices before and after the round‑trip are identical.

Beyond trees, the authors extend the methodology to arbitrary unweighted graphs. First, they compute the all‑pairs shortest‑path distance matrix (using Floyd‑Warshall or equivalent). This distance matrix is then interpreted as a fully connected weighted graph where each edge weight equals the shortest‑path length in the original graph. Applying the same agglomerative clustering to this weighted representation yields a dendrogram whose branches correspond to the community structure of the original network. In this way, community detection becomes a natural by‑product of the tree‑construction process, and the resulting tree provides a clear hierarchical view of the network’s modular organization.

The paper highlights several practical implications. By converting trees into networks, a wealth of off‑the‑shelf complex‑network metrics—clustering coefficient, degree assortativity, betweenness centrality, etc.—can be directly applied to tree data, offering new quantitative descriptors for phylogenetic trees, vascular branching, or any hierarchical system. Conversely, representing a graph as a tree exposes its modular hierarchy, facilitating intuitive visualization and enabling the identification of missing intra‑community links by comparing the original unweighted graph with the fully connected weighted version derived from the tree‑network‑tree transformation.

Experimental illustrations include a simple synthetic tree transformed into a TAN, where standard network measures reproduce the tree’s branching pattern, and an unweighted graph with two clear communities that, after distance‑based weighting and agglomerative clustering, emerges as a binary tree separating the two groups. The authors note that the forward tree‑to‑network mapping runs in linear time O(N), while the distance‑based conversion for arbitrary graphs is O(N³) due to the all‑pairs shortest‑path computation. Nevertheless, they report successful application to networks of a few thousand nodes and suggest that parallel or approximate distance algorithms could scale the approach to larger systems.

In conclusion, the work establishes a perfect duality between trees and weighted networks, showing that each can be faithfully represented in terms of the other. This duality opens new avenues for both theoretical analysis—linking hierarchical models with graph theory—and practical applications such as enhanced community detection, hierarchical visualization, and enriched characterization of biological or technological trees using complex‑network tools. Future directions proposed include exploring alternative distance metrics, benchmarking computational performance on large real‑world datasets, and integrating the mapping into novel clustering algorithms.