How to Improve the Stability of Clustering Results with VGAE?

[heheyiyi]

Thank you for sharing your code and great work! I have been running the code and using the VGAE to extract features for clustering. However, I’ve noticed that the clustering results are inconsistent across different runs—sometimes the results are good, and other times they are less optimal. I assume this variation is due to the randomness involved in the VGAE feature extraction process.

I would like to ask for your advice on the following:

1、How can I achieve more stable and consistent clustering results? Are there any techniques to minimize the variability in the features extracted by the VGAE?

2、Could adjusting the model or clustering algorithm parameters, or applying some kind of post-processing (e.g., dimensionality reduction or regularization), help to improve the stability of the clustering results?