Eigenvectors embedding

One simple way of reducing the size of the matrix is to decompose it into eigenvectors, and use only a reduced number of these vectors as embedding.

An example of such graph representation can be seen when using graph positioning. Indeed, drawing a graph on a two-dimensional plane is a type of embedding. One of the positioning techniques that can be found in networkx is called spectral_layout and consists of decomposing the Laplacian matrix into its eigenvectors.