The unsupervised graph embedding roadmap
Unsupervised machine learning involves algorithms that can be trained without the need for manually annotated data, making them especially valuable for identifying hidden structures and relationships in complex graph networks. Most of these models rely only on information in the adjacency matrix and the node features, without any knowledge of the downstream machine learning task.
How can this be done? One of the most common approaches is to learn embeddings that preserve the graph structure. The learned representation is usually optimized so that it can be used to reconstruct the pair-wise node similarity, for example, the adjacency matrix. These techniques bring an important feature: the learned representation can encode latent relationships among nodes or graphs, allowing us to discover hidden and complex novel patterns.
Many algorithms have been developed in relation to unsupervised graph machine learning techniques. However, as...
