Controllable graph diffusion for community search

Tiesunlong Shen

doi:10.1016/j.inffus.2026.104655

Controllable graph diffusion for community search

2026-07-28

0

PRE

AI

DOI：10.1016/j.inffus.2026.104655

Original

Original request for help

Save

Abstract

En 中文

• CGD is a controllable graph diffusion model that recasts community search as a community generation task. • A discrete diffusion process preserves the original graph topology, which continuous graph diffusion models cannot retain. • A copilot network handles attributed community search without retraining the full model. • Experiments on nine real-world graph datasets show CGD outperforms state-of-the-art community search baselines.

Keywords:

Diffusion

Graph pattern recognition

Graph neural network

Controllable graph diffusion for community search

Abstract

Journal

Organization