Diversity Curves for Graph Representation Learning
About
Graph-level representations are crucial tools for characterising structural differences between graphs. However, comparing graphs with different cardinalities, even when sampled from the same underlying distribution, remains challenging. Unsupervised tasks in particular require interpretable, scalable, and reliable size-aware graph representations. Our work addresses these issues by tracking the structural diversity of a graph across coarsening levels. The resulting graph embeddings, which we denote diversity curves, are interpretable by construction, efficient, and directly comparable across coarsening hierarchies. Specifically, we track the spread of graphs, a novel isometry invariant that is inherently well-suited for encoding the metric diversity and geometry of graphs. We utilise edge contraction coarsening and prove that this improves expressivity, thus leading to more powerful graph-level representations than structural descriptors alone. Demonstrating their utility over a range of baseline methods in practice, we use diversity curves to (i) cluster and visualise simulated graphs across varying sizes, (ii) distinguish the geometry of single-cell graphs, (iii) compare the structure of molecular graph datasets, and (iv) characterise geometric shapes.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Graph Distribution Classification and Clustering | ER, RP, SBM, RG graph distributions | Accuracy90.4 | 31 | |
| Graph classification (trajectory- vs cluster-like) | Single-cell graphs All Graphs (full set) | Accuracy76.6 | 28 | |
| Graph classification (trajectory- vs cluster-like) | Single-cell graphs Gold subset | Accuracy86 | 27 | |
| Graph Expressivity | BREC | Basic Expressivity Score60 | 26 | |
| Graph Distribution Classification | Random Graph Models ER, RP, RG, SBM | Accuracy90 | 8 | |
| Graph Parameter Classification | Erdős-Rényi (ER) random graph model | Accuracy100 | 8 | |
| Graph Parameter Classification | Random Partition (RP) | Accuracy98 | 8 | |
| Graph Parameter Classification | Random Geometric (RG) Graph | Accuracy97 | 8 | |
| Graph Parameter Classification | Stochastic block model (SBM) | Accuracy86 | 8 | |
| Binary Graph Classification | All 169 Graphs (5-fold stratified CV) | Accuracy (Test)75.9 | 6 |