There are various approaches to solve graph classification problem. Graph embedding is one of them. This GitHub repository provide a from scratch implementation of Spectral Feature [1] as one of embedding method available. The calculation of this feature involves adjacency matrix and eigenvalue of Laplacian matrix of the graph.
[1] N. de Lara and E. Pineau, "A Simple Baseline Algorithm for Graph Classification" In Relational Representation Learning, NIPS 2018 Workshop, Montréal, Canada.
The official implementation from the author can be accessed here.