LARW: Network Representation Learning Algorithm Based On Long Anonymous Random Walks

Abstract

Network Representation Learning (NRL) plays an important role in network analysis and aims to represent complex networks more concisely by transforming nodes into low-dimensional vectors. Network representation learning has become the focus of increasing research interest in academia and industry. Much of the work in one direction is based on learning network features based on random walk derived models. However, it is difficult to extract effective features in the face of long wandering sequences. Therefore, we propose a dynamic network characterization method based on Long Anonymous Random Walks(LARW). LARW incorporates the latest long series prediction method Informer, which allows more feature information to be retained. The model parameters are optimized in the process of comparing with the actual results, thus making the node embedding more predictive and causal. In our experiments, we compare our model with the existing NRL model on four real-world datasets. The experimental results show that LARW achieves superior results in tasks such as node classification and link prediction.