What are node embeddings?
Node embeddings are vectors (tensors) of numbers that are obtained by a process called embedding. So what's this embedding process? the way i think of it is a process of mapping something complicated (a graph, a document, an image) into simple fixed length vectors while retain its key features. For an example, in graph, it would mean to translate the whole or parts of graph in N-dimensional space. Take a look at the example below. In this example, all the nodes are mapped in 2-dimensional space.
What do we use node embeddings for?
- Node Classification
- Link Prediction
- Clustering
Top comments (0)