![]() The rest of this paper is structured as follows. To the best of our knowledge, this work is the first survey focused on this subject. In this survey, we try to bring an overview of works related to bipartite graph embeddings and gather a list of tools available to explore this area. Nevertheless, representation learning for bipartite graphs has only received little attention, with a few papers dedicated to it. 2017).īipartite graphs have a particular importance since many real-world applications such as topic modeling, medical diagnosis or recommendation can be modeled as computing on bipartite graphs (see Table 2 for more examples of application). 2021), or bipartite networks (Stauffer et al. ![]() 2021), heterogeneous networks (Yang et al. Consequently, a variety of tailor-made methods have been created to investigate specific network types, e.g., homogeneous networks (Cai et al. In particular, recent years have seen the emergence of several sub-domains of graph embeddings with a particular focus on the considered network type. With quantities of data exploding, this area of research has attracted a strong interest from the scientific community and inspired numerous works. Such representations require less effort to be handled hence, they can be used as features for common tasks on graphs and are directly usable by various downstream machine learning algorithms. Precisely, the purpose of embeddings is to find a mapping function which associates a low-dimensional latent representation to each node in the network. The main principle here is to reduce the dimension of the network in an embedding space while preserving major network characteristics (structure, nodes’ proximity, edges). Therefore, it can be intractable to perform complex inference procedures on the entire network.Ī common way to bypass this limitation is network representation learning, also known as network embedding (Arsov and Mirceva 2019). In the era of big data, information networks can contain billions of nodes and edges. However, this advantage may turn as a downside because of the processing capacities that they require. Networks offer a rich structure that can model complex relationships. They are used in topics as diverse as computational biology, recommendation systems, finance or social networks (Hegeman and Iosup 2018). ![]() Networks come as a natural way to model a diverse set of real-world information.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |