Improving Skip-Gram based Graph Embeddings via Centrality-Weighted Sampling

Abstract

Network embedding techniques inspired by word2vec represent an e ective unsuper- vised relational learning model. Commonly, by means of a Skip-Gram procedure, these techniques learn low dimensional vector representations of the nodes in a graph by sam- pling node-context examples. Although many ways of sampling the context of a node have been proposed, the e ects of the way a node is chosen have not been analyzed in depth. To ll this gap, we have re-implemented the main four word2vec inspired graph embedding techniques under the same framework and analyzed how di erent sampling distributions a ects embeddings performance when tested in node classi cation problems. We present a set of experiments on di erent well known real data sets that show how the use of popular centrality distributions in sampling leads to improvements, obtaining speeds of up to 2 times in learning times and increasing accuracy in all cases.