Publications

Retrofitting structural graph embeddings with node attribute information
Retrofitting structural graph embeddings with node attribute information

Representation learning for graphs has attracted increasing attention in recent years. In this paper, we define and study a new problem of learning attributed graph embeddings. Our setting considers how to update existing node representations from structural graph embedding methods when some additional node attributes are given. To this end, we propose Graph Embedding RetroFitting (GERF), a method that delivers a compound node embedding that follows both the graph structure and attribute space similarity. Unlike other attributed graph embedding methods, GERF is a novel representation learning method that does not require recalculation of the embedding from scratch but rather uses existing ones and retrofits the embedding according to neighborhoods defined by the graph structure and the node attributes space. Moreover, our approach keeps the same embedding space all the time and allows comparing the positions of embedding vectors and quantifying the impact of attributes on the representation update. Our GERF method updates embedding vectors by optimizing the invariance loss, graph neighbor loss, and attribute the neighbor loss to obtain high-quality embeddings. Experiments on WikiCS, Amazon-CS, Amazon-Photo, and Coauthor-CS datasets demonstrate that our proposed algorithm receives similar results compared to other state-of-the-art attributed graph embedding models despite working in retrofitting manner.

FILDNE: A Framework for Incremental Learning of Dynamic Networks Embeddings
FILDNE: A Framework for Incremental Learning of Dynamic Networks Embeddings

Representation learning on graphs has emerged as a powerful mechanism to automate feature vector generation for downstream machine learning tasks. The advances in representation on graphs have centered on both homogeneous and heterogeneous graphs, where the latter presenting the challenges associated with multi-typed nodes and/or edges. In this paper, we consider the additional challenge of evolving graphs. We ask the question of whether the advances in representation learning for static graphs can be leveraged for dynamic graphs and how? It is important to be able to incorporate those advances to maximize the utility and generalization of methods. To that end, we propose the Framework for Incremental Learning of Dynamic Networks Embedding (FILDNE), which can utilize any existing static representation learning method for learning node embeddings while keeping the computational costs low. FILDNE integrates the feature vectors computed using the standard methods over different timesteps into a single representation by developing a convex combination function and alignment mechanism. Experimental results on several downstream tasks, over seven real-world datasets, show that FILDNE is able to reduce memory (up to 6x) and computational time (up to 50x) costs while providing competitive quality measure gains (e.g., improvements up to 19 pp AUC on link prediction and up to 33 pp mAP on graph reconstruction) with respect to the contemporary methods for representation learning on dynamic graphs.