ripple2vec: Node Embedding with Ripple Distance of Structures

Abstract Graph is a generic model of various networks in real-world applications. And, graph embedding aims to represent nodes (edges or graphs) as low-dimensional vectors which can be fed into machine learning algorithms for downstream analysis tasks. However, existing random walk-based node methods often map some with (dis)similar local structures (near) far vectors. To overcome this issue, paper proposes implement by constructing context via new defined ripple distance over vectors, whose components are the hitting times fully condensed neighborhoods and thus characterize their pure quantities. The able capture (dis)similarities nodes’ neighborhood satisfies triangular inequality. neighbors each distance, makes short walks from given only visit its similar original graph. This property guarantees that proposed method, named $$\mathsf {ripple2vec}$$ ripple2vec , (far) near Experimental results on real datasets, where labels mainly related structures, show behave better than those state-of-the-art methods, clustering classification, competitive other link prediction.