The article presents a novel approach to multi-relational clustering, which is a challenging task due to the diverse semantic information in multi-layer graphs. The authors propose a graph filter based on the theoretical analysis of Barlow Twins, a popular correlation-based objective function. They find that input with a negative semi-definite inner product provides a lower bound for Barlow Twins loss, which prevents it from reaching a better solution. To overcome this, they learn a filter that yields an upper bound for Barlow Twins. The proposed method shows state-of-the-art performance on four benchmark datasets.
Publication date: 22 Dec 2023
Project Page: https://github.com/XweiQ/BTGF
Paper: https://arxiv.org/pdf/2312.14066
