V. M. Viswanatham

ADAPT-Net: an adaptive dynamic attention and persistence-aware transformer for overlapping community detection in complex social networks

The proposed framework integrates adaptive event-driven graph partitioning, multiscale graph attention encoding, learnable node-centric scale fusion, cross-snapshot link-aware temporal attention, and density-preserving overlap regularization into an adaptive dynamic attention and persistence-aware transformer for overlapping community detection in complex social networks.