To address the limitations of existing methods in extracting snapshot features and adapting to varying network scales, this paper proposes a novel, unsupervised, end-to-end framework called the Variational Graph Gaussian Mixture model (VGGM) for change point detection in dynamic networks.
VGGM effectively extracts snapshot embeddings by combining a Variational Graph Auto-Encoder (VGAE) with a dedicated readout function, and it automates the detection of change points using a Gaussian Mixture Model (GMM).
Through joint iterative training, this approach models dynamic networks using a Mixture-of-Gaussians prior, demonstrating superior performance over current state-of-the-art methods on both real-world and synthetic datasets
To address the limitations of existing methods in extracting snapshot features and adapting to varying network scales, this paper proposes a novel, unsupervised, end-to-end framework called the Variational Graph Gaussian Mixture model (VGGM) for change point detection in dynamic networks.
VGGM effectively extracts snapshot embeddings by combining a Variational Graph Auto-Encoder (VGAE) with a dedicated readout function, and it automates the detection of change points using a Gaussian Mixture Model (GMM).
Through joint iterative training, this approach models dynamic networks using a Mixture-of-Gaussians prior, demonstrating superior performance over current state-of-the-art methods on both real-world and synthetic datasets