Nane: A Node2vec Extension for Attributed Network Embedding

Traditional network representation learning methods focus solely on the network’s topology, ignoring other sources of information that could improve the learning process. On the other hand, attributed networks incorporate additional contextual information in the form of node attributes, which can lead to more accurate representations of nodes in the network. The proposed approach aims to map the network onto a low-dimensional space that effectively captures the interaction between the two sources of information. In this study, we present an extension of the node2vec algorithm, called Node2vec Attributed Network Embedding that incorporates both network topology and node attributes to learn network embeddings. We evaluate the performance of Node2vec Attributed Network Embedding against other state-of-the-art methods for node classification and link prediction tasks on real-world datasets, demonstrating that Node2vec Attributed Network Embedding outperforms other methods and highlighting the importance of incorporating diverse feature types for network representation learning. Our study provides valuable insights into the challenges of representing network data for machine learning tasks. It proposes a practical approach for incorporating structural and attribute information into the network embedding process.