Abstract
Disconnected graphs are very common in the real
world. However, most existing methods for graph
similarity focus on connected graph. In this paper,
we propose an effective approach for measuring the
similarity of disconnected graphs. By embedding
connected subgraphs with graph kernel, we obtain
the feature vectors in low dimensional space. Then,
we match the subgraphs and weigh the similarity of
matched subgraphs. Finally, an intuitive example
shows the feasibility of the method