Recently, team members have received inquiries from researchers on social media/knowledge platforms regarding whether the CDNMF algorithm is suitable for directed graphs.
In fact, our CDNMF algorithm is theoretically adaptable to community detection in directed graphs. It only requires inputting the adjacency matrix/feature matrix of the directed graph and making appropriate adjustments in the graph regularization part, while the rest can remain unchanged. Specifically, when calculating the Laplacian matrix, you can use:
L = D - (A + A^T) , or- the Laplacian matrix for directed graphs, such as the one referenced in:
Laplacians and the Cheeger Inequality for Directed Graphs.
Additionally, if you wish to better capture directional information, the model can be further extended.
We welcome interested contributors to submit relevant Pull Requests (PRs) and work together to improve the CDNMF algorithm. We are happy to review and merge contributions — thank you very much! ~
Recently, team members have received inquiries from researchers on social media/knowledge platforms regarding whether the CDNMF algorithm is suitable for directed graphs.
In fact, our CDNMF algorithm is theoretically adaptable to community detection in directed graphs. It only requires inputting the adjacency matrix/feature matrix of the directed graph and making appropriate adjustments in the graph regularization part, while the rest can remain unchanged. Specifically, when calculating the Laplacian matrix, you can use:
L = D - (A + A^T), orLaplacians and the Cheeger Inequality for Directed Graphs.
Additionally, if you wish to better capture directional information, the model can be further extended.
We welcome interested contributors to submit relevant Pull Requests (PRs) and work together to improve the CDNMF algorithm. We are happy to review and merge contributions — thank you very much! ~