FODGE - Fast Online Dynamic Graph Embedding

Graph embedding algorithms (GEA) project each vertex in a graph to a real-valued vector. Dynamic GEA (DGEA) are used to project dynamic graphs, where vertices and edges can appear and disappear. Such graphs are mostly divided into snapshots. Current DGEAs are often offline, and computationally expensive, and most do not ensure a slow change in vertices projection.We propose here FODGE, a novel DGEA algorithm to gradually shift the projection of vertices whose first and second neighbors changed. FODGE optimizes CPU and memory efficacy by initially projecting the graph's densest K-core using any existing global optimization and then projecting the periphery of the graph using a local approximation. FODGE then smoothly updates the projection of all vertices, through an iterative local update rule. As such it can be applied to extremely large dynamic graphs over long periods.We show that FODGE is faster than current algorithms, more accurate in an auxiliary task of link prediction. and it ensures a limited difference in vertex positions between consecutive time points. FODGE is highly modular and can be combined with any static projection, including graph convolutional networks, and has a few hyperparameters to tune. The code is available at https://github.com/unknownuser13570/FODGE in GIT.