Graph neural network-based interactive clustering enhanced by human knowledge

doi:10.1016/j.asoc.2025.113595

Graph neural network-based interactive clustering enhanced by human knowledge

2025-08-20

0

PRE

AI

DOI：10.1016/j.asoc.2025.113595

Original

Original request for help

Save

Abstract

En 中文

• Guided Human-in-the-Loop Interaction System. A composite rank metric and kNN-based algorithm streamline cluster-to-sample adjustments, reducing cognitive load and errors. • Graph-Driven Preference Modeling. A latent graph algorithm dynamically encodes user-defined relationships using edge density and modularity, bridging human intuition with algorithmic generalization. • Scalable Semi-Supervised Clustering. A multi-constraint GNN propagates sparse user annotations on a latent graph, enabling large-scale, semi-supervised classification without retraining.

Keywords:

Guided Human-in-the-Loop Interaction

Graph-Driven Preference Modeling

Scalable Semi-Supervised Clustering

kNN-based Algorithm

Multi-constraint GNN

Graph neural network-based interactive clustering enhanced by human knowledge

Abstract

Journal

Organization