Modified SkipGram Negative Sampling Model for Faster Convergence of Graph Embedding

Graph embedding techniques have been introduced in recent years with the aim of mapping graph data into low-dimensional vector spaces, so that conventional machine learning methods can be exploited. In particular, in the DeepWalk model, truncated random walks are employed in random walk-based approaches to capture structural links-connections between nodes. The SkipGram model is then applied to the truncated random walks to compute the embedded nodes. In this work, the proposed DeepWalk model provides a faster convergence speed than the standard one by introducing a new trainable parameter in the model. Furthermore, experimental results on real-world datasets show that the performance in downstream community detection and link prediction task is improved by using the proposed DeepWalk model.