SimRank on Data Streams

This paper introduces a novel approach that integrates machine learning techniques to optimize SimRank convergence on streaming data. Our approach utilizes the structural properties of the graph, such as edge features, to train a linear regression model that predicts SimRank values for newly added nodes. These predictions serve as effective initializations, ensuring faster convergence of the SimRank algorithm. By avoiding full recomputation and providing a new direction for research, our method offers a promising solution for similarity computation in evolving networks.