A Scalable Unsupervised Framework for Comparing Graph Embeddings
- 117 Downloads
Graph embedding is a transformation of vertices of a graph into a set of vectors. A good embedding should capture the graph topology, vertex-to-vertex relationship, and other relevant information about the graph, its subgraphs, and vertices. If these objectives are achieved, an embedding is a meaningful, understandable, and often compressed representations of a network. Unfortunately, selecting the best embedding is a challenging task and very often requires domain experts.
In the recent paper , we propose a “divergence score” that can be assigned to embeddings to help distinguish good ones from bad ones. This general framework provides a tool for an unsupervised graph embedding comparison. The complexity of the original algorithm was quadratic in the number of vertices. It was enough to show that the proposed method is feasible and has practical potential (proof-of-concept). In this paper, we improve the complexity of the original framework and design a scalable approximation algorithm. Moreover, we perform some detailed quality and speed benchmarks.
KeywordsGraph embedding Geometric Chung-Lu Model
- 1.Kamiński, B., Prałat, P., Théberge, F.: An unsupervised framework for comparing graph embeddings. J. Complex Networks, in press. 27 pGoogle Scholar
- 7.McInnes, L., Healy, J., Melville, J.: UMAP: uniform manifold approximation and projection for dimension reduction. pre-print arXiv:1802.03426 (2018)
- 10.Leskovec, J., Krevl, A.: SNAP datasets: Stanford large network dataset collection. http://snap.stanford.edu/data
- 11.Kamiński, B., Prałat, P., Théberge, F.: Artificial benchmark for community detection (ABCD) – fast random graph model with community structure, pre-print arXiv:2002.00843 (2020)
- 14.Antelmi, A., et al.: Analyzing, exploring, and visualizing complex networks via hypergraphs using SimpleHypergraphs.jl. Internet Math. (2020). 32 pGoogle Scholar