Graph models, like other machine learning models, have implicit and explicit biases built-in, which often impact performance in nontrivial ways. The model’s faithfulness is often measured by comparing the newly generated graph against the source graph using any number of graph properties. Therefore, differences in the size or topology of the generated graph indicate a loss in the model. Yet, in many systems, errors encoded in loss functions are subtle and not well understood. In the present work, we introduce theInfinity Mirrortest for analyzing the robustness of graph models. This straightforward stress test works by repeatedly fitting a model to its outputs. A hypothetically perfect graph model would have no deviation from the source graph; however, a model’s implicit biases and assumptions are exaggerated by the Infinity Mirror test, exposing potential previously obscured issues. Through an analysis of thousands of experiments on synthetic and real-world graphs, we show that several conventional graph models degenerate in exciting and informative ways. We believe that the observed degenerative patterns are clues to the future development of better graph models.

Recent work at the intersection of formal language theory and graph theory has explored graph grammars for graph modeling. However, existing models and formalisms can only operate on homogeneous (i.e., untyped or unattributed) graphs. We relax this restriction and introduce the Attributed Vertex Replacement Grammar (AVRG), which can be efficiently extracted from heterogeneous (i.e., typed, colored, or attributed) graphs. Unlike current state-of-the-art methods, which train enormous models over complicated deep neural architectures, the AVRG model is unsupervised and interpretable. It is based on context-free string grammars and works by encoding graph rewriting rules into a graph grammar containing graphlets and instructions on how they fit together. We show that the AVRG can encode succinct models of input graphs yet faithfully preserve their structure and assortativity properties. Experiments on large real-world datasets show that graphs generated from the AVRG model exhibit substructures and attribute configurations that match those found in the input networks.

We generalize triadic closure, along with previous generalizations of triadic closure, under an intuitive umbrella generalization: the Subgraph-to-Subgraph Transition (SST). We present algorithms and code to model graph evolution in terms of collections of these SSTs. We then use the SST framework to create link prediction models for both static and temporal, directed and undirected graphs which produce highly interpretable results. Quantitatively, our models match out-of-the-box performance of state of the art graph neural network models, thereby validating the correctness and meaningfulness of our interpretable results.

An enormous amount of real-world data exists in the form of graphs. Oftentimes, interesting patterns that describe the complex dynamics of these graphs are captured in the form of frequently reoccurring substructures. Recent work at the intersection of formal language theory and graph theory has explored the use of graph grammars for graph modeling and pattern mining. However, existing formulations do not extract meaningful and easily interpretable patterns from the data. The present work addresses this limitation by extracting a special type of vertex replacement grammar, which we call a KT grammar, according to the Minimum Description Length (MDL) heuristic. In experiments on synthetic and real-world datasets, we show that KT-grammars can be efficiently extracted from a graph and that these grammars encode meaningful patterns that represent the dynamics of the real-world system.

One of the principal goals of graph modeling is to capture the building blocks of network data in order to study various physical and natural phenomena. Recent work at the intersection of formal language theory and graph theory has explored the use of graph grammars for graph modeling. However, existing graph grammar formalisms, like Hyperedge Replacement Grammars, can only operate on small tree-like graphs. The present work relaxes this restriction by revising a different graph grammar formalism called Vertex Replacement Grammars (VRGs). We show that a variant of the VRG called Clustering-based Node Replacement Grammar (CNRG) can be efficiently extracted from many hierarchical clusterings of a graph. We show that CNRGs encode a succinct model of the graph, yet faithfully preserves the structure of the original graph. In experiments on large real-world datasets, we show that graphs generated from the CNRG model exhibit a diverse range of properties that are similar to those found in the original networks.